Studies in Computational Intelligence 929

Jongbae Kim
Roger Lee Editors

Data Science
and Digital
Transformation
in the Fourth
Industrial
Revolution
Studies in Computational Intelligence

Volume 929

Series Editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
The series “Studies in Computational Intelligence” (SCI) publishes new develop-
ments and advances in the various areas of computational intelligence—quickly and
with a high quality. The intent is to cover the theory, applications, and design
methods of computational intelligence, as embedded in the fields of engineering,
computer science, physics and life sciences, as well as the methodologies behind
them. The series contains monographs, lecture notes and edited volumes in
computational intelligence spanning the areas of neural networks, connectionist
systems, genetic algorithms, evolutionary computation, artificial intelligence,
cellular automata, self-organizing systems, soft computing, fuzzy systems, and
hybrid intelligent systems. Of particular value to both the contributors and the
readership are the short publication timeframe and the world-wide distribution,
which enable both wide and rapid dissemination of research output.
Indexed by SCOPUS, DBLP, WTI Frankfurt eG, zbMATH, SCImago.
All books published in the series are submitted for consideration in Web of Science.

More information about this series at http://www.springer.com/series/7092

Jongbae Kim Roger Lee
•

Editors

Data Science and Digital

Transformation in the Fourth
Industrial Revolution

123
Editors
Jongbae Kim Roger Lee
Startup Support Foundation ACIS International
Soongsil University Mount Pleasant, MI, USA
Seoul, Korea (Republic of)

ISSN 1860-949X ISSN 1860-9503 (electronic)

Studies in Computational Intelligence
ISBN 978-3-030-64768-1 ISBN 978-3-030-64769-8 (eBook)
https://doi.org/10.1007/978-3-030-64769-8
© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2021
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

The purpose of International Semi-Virtual Workshop on Data Science and Digital

Transformation in the Fourth Industrial Revolution (DSDT 2020) held on October
17, 2020, Soongsil University, Seoul, Korea, is aimed at bringing together
researchers and scientists, businessmen and entrepreneurs, teachers and students to
discuss the numerous fields of computer science and to share ideas and information
in a meaningful way. This workshop on data science and digital transformation in
the Fourth Industrial Revolution will discuss the wide range of issues with sig-
nificant implications, from data warehousing, to data mining, online analytical
processing and reporting, data quality assessment, data-driven business models,
data privacy and security issues, etc.
This publication captures 17 of the workshop’s most outstanding papers. The
selected papers have not been published in the workshop proceedings or elsewhere,
but only in this book. We await the important contributions that we know these
authors will bring to the field.
In Chapter “A Study on the Intention to Use Korean Telemedicine Services:
Focusing on the UTAUT2 Model,” Harim Byun and Jongwoo Park investigate the
factors influencing the intention to use Korean telemedicine services from the
perspective of potential overseas customers. This study has tested the influence
relationship by reconstructing the UTAUT2 model to be suitable for the studies on
the acceptance of telemedicine services.
In Chapter “A Structural Relationship Between Environmental Uncertainty,
Dynamic Capability, and Business Performance in a Smart Supply Chain
Environment,” Yongmuk Kim and Jongwoo Park tested their hypothesis using
structural equation model analysis to verify the structural relationship between
dynamic capability and business performance in the smart supply chain environ-
ment of Korean small and medium manufacturers. They confirmed through this that
dynamic capability and environmental uncertainty in a smart supply chain envi-
ronment have a partially significant effect on business performance and that man-
ufacturers must strive to implement direct, systematic policies and improve
dynamic capability to respond to an uncertain business environment.

v
vi Preface

In Chapter “Analysis of the Relative Importance of Coaching Service Quality,”

Jinyoung Yang, Hyoboon Wang and Donghyuk Jo analyze the relative importance
between determinants and factors of coaching service quality. Result of this study
may be used as basic data in establishing an academic knowledge system of
coaching and expanding the base for coaching.
In Chapter “Determinants of Technology-Based Self-Service Acceptance,”
Seulki Lee and Donghyuk Jo explore the factors that effects the acceptance
intention of technology-based self-service of food and beverage store customers.
The results of this study have academic implications for investigating the ante-
cedent variables to the acceptance intention of technology-based self-service in
uncertain environments. In practice, it will contribute to decision-making and
business strategy formulation for the post-corona era.
In Chapter “Pre-verification of Data in Electronic Trade Blockchain Platform,”
Saeyong Oh, Sanghyun Cho, Sunghwa Han and Gwangyong Gim propose an
architecture that can verify the integrity of the blockchain-based electronic docu-
ment platform by presenting a pre-validation method for electronic documents.
In Chapter “Study on Security and Privacy of E-Government Service,”
Sanghyun Cho, Saeyong Oh, Hogun Rou and Gwangyong Gim analyzed prior
studies and cases of e-government in order to derive the factors influencing the
continuous use of e-government services provided by central government depart-
ments, local governments and public institutions, and to establish the causal rela-
tionship between each factor.
In Chapter “Applying Hofstede’s Culture Theory in the Comparison between
Vietnam and Korean E-government Adoption,” Hung-Trong Van, Simon Gim,
Euntaek Lim and Thi-Thanh-Thao Vo find the factors affecting the intention to use
e-government by comparing Korea and Vietnam citizens’ behavior based on TAM
and IS success model.
In Chapter “Study on Business Strategy Quantification using Topic Modeling
and Word Embedding: Focusing on ‘Virtual Reality’ and ‘Augmented Reality’,”
Siyoung Lee, Sungwoong Seo, Hyunjae Yoo and Gwangyong Gim analyzed arti-
cles related to “virtual reality” and “augmented reality,” the main elements of the
Fourth Industrial Revolution, and reviewed the views that the media has by using
text mining techniques. This study showed that topics derived from topic modeling
can be expressed in strategic space both nominally and visually.
In Chapter “3D Printing Signboard Production Using 3D Modeling Design,”
Jungkyu Moon and Deawoo Park proposed 3D printing signboard production using
3D modeling design. The signboard studied in this paper has the advantages of long
life and low failure rate, and the disadvantage of high cost is to be solved by using
fourth industrial technology.
In Chapter “AI-Based 3D Food Printing Using Standard Composite Materials,”
Hyunju Yoo and Daewoo Park proposed AI-based 3D food printing using standard
composite materials. This study overcome the limitation of 3D food printing, which
cannot have standardized design values for printability due to the variety of food
raw materials and materials added to improve printability.
Preface vii

In Chapter “Telemedicine AI App for Prediction of Pets Joint Diseases,” Suyeon

Han and Deawoo Park proposed a smartphone app based on 5G communication that
can diagnose the presence and possibility of bone joint disease in companion animals.
In Chapter “Design of Artificial Intelligence for Smart Risk Pre-review System
at the KC EMC,” Youngjoo Oh and Deawoo Park proposed an AI deep learning
system that provides a complementary measure and provide it in the form of a cloud
computing platform. And they proposed a semi-automatic AI connection system
that can perform KC conformity assessment for electric and electronic products to
be sold in Korea.
In Chapter “AI Analysis of Illegal Parking Data at Seocho City,” Donghyun Lim
and Deawoo Park proposed an AI machine learning system that links the vehicle's
number recognition algorithm for illegal parking. And they suggest an advanced
system that analyzes the status of illegal parking and stopping judgment in Seocho
City Office, where big data and AI are connected using spatial information and AI.
In Chapter “Regularized Categorical Embedding for Effective Demand
Forecasting of Bike Sharing System,” Sangho Ahn, Hansol Ko and Juyoung
Kang propose a regularized categorical embedding methodology that can learn not
only the station-centric model but also the overall trend to predict the rental demand
for shared bicycles. The proposed methodology is expected to be an effective
prediction for station-centric models with unbalanced performance distribution and
the prediction of rental demand for shared bicycles.
In Chapter “Development of a Model for Predicting the Demand for Bilingual
Teachers in Elementary Schools to Support Multicultural Families—Based on
NEIS Data,” Jinmyung Choi and Dooyeon Kim developed a test model using big
data to analyze the data of school sites and data held by administrative agencies and
applied the actual data to this model to analyze the policy’s appropriateness and
effectiveness. This study’s results can help establish and implement policies in
various educational fields through big data analysis.
In Chapter “Research on Implementation of User Authentication Based on
Gesture Recognition of Human,” Jungseon Oh, Joongyoung Choi, Kwansik Moon
and Kyoungho Lee propose a method to implement such biometric authentication
based on gestures (dynamic characteristics) and body characteristics (static char-
acteristics) and suggest ways to use them.
In Chapter “AI TTS Smartphone App for Communication of Speech Impaired
People,” Hanyoung Lee and Deawoo Park proposed AI TTS smartphone app for
communication of speech impaired people. This study will be used as a means of
transmitting information to the hearing impaired in the untact era.
It is our sincere hope that this volume provides stimulation and inspiration, and
that it will be used as a foundation for works to come.

Seoul, Korea (Republic of) Jongbae Kim

August 2020
Contents

A Study on the Intention to Use Korean Telemedicine Services:

Focusing on the UTAUT2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Harim Byun and Jongwoo Park
A Structural Relationship Between Environmental Uncertainty,
Dynamic Capability, and Business Performance in a Smart
Supply Chain Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Yongmuk Kim and Jongwoo Park
Analysis of the Relative Importance of Coaching Service Quality . . . . . 25
Jinyoung Yang, Hyoboon Wang, and Donghyuk Jo
Determinants of Technology-Based Self-Service Acceptance . . . . . . . . . . 39
Seulki Lee and Donghyuk Jo
Pre-verification of Data in Electronic Trade Blockchain Platform . . . . . 53
Saeyong Oh, Sanghyun Cho, Sunghwa Han, and Gwangyong Gim
Study on Security and Privacy of E-Government
Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Sanghyun Cho, Saeyong Oh, Hogun Rou, and Gwangyong Gim
Applying Hofstede’s Culture Theory in the Comparison Between
Vietnam and Korean E-Government Adoption . . . . . . . . . . . . . . . . . . . 81
Hung-Trong Van, Simon Gim, Euntaek Lim, and Thi-Thanh-Thao Vo
Study on Business Strategy Quantification Using Topic Modeling
and Word Embedding: Focusing on ‘Virtual Reality’
and ‘Augmented Reality’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Siyoung Lee, Sungwoong Seo, Hyunjae Yoo, and Gwangyong Gim
3D Printing Signboard Production Using 3D Modeling Design . . . . . . . 109
Jungkyu Moon and Deawoo Park

ix
x Contents

AI-Based 3D Food Printing Using Standard Composite Materials . . . . . 123

Hyunju Yoo and Daewoo Park
Telemedicine AI App for Prediction of Pets Joint Diseases . . . . . . . . . . 137
Suyeon Han and Deawoo Park
Design of Artificial Intelligence for Smart Risk Pre-review System
at the KC EMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Youngjoo Oh and Deawoo Park
AI Analysis of Illegal Parking Data at Seocho City . . . . . . . . . . . . . . . . 165
Donghyun Lim and Deawoo Park
Regularized Categorical Embedding for Effective Demand
Forecasting of Bike Sharing System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Sangho Ahn, Hansol Ko, and Juyoung Kang
Development of a Model for Predicting the Demand for Bilingual
Teachers in Elementary Schools to Support Multicultural
Families—Based on NEIS Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Jinmyung Choi and Dooyeon Kim
Research on Implementation of User Authentication Based on Gesture
Recognition of Human . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Jungseon Oh, Joongyoung Choi, Kwansik Moon, and Kyoungho Lee
AI TTS Smartphone App for Communication of Speech
Impaired People . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Hanyoung Lee and Deawoo Park

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

Contributors

Sangho Ahn Department of e-Business, Ajou University, Suwon, Republic of

Korea
Harim Byun Soongsil University, Seoul, South Korea
Sanghyun Cho Department of IT Policy and Management, Soongsil University,
Seoul, South Korea
Jinmyung Choi Department of Convergence IT, Konyang University,
Chungchungnam-Do, Republic of Korea
Joongyoung Choi Defense Information System Management Group, KIDA,
Seoul, Republic of Korea
Gwangyong Gim Department of IT Policy and Management, Soongsil University,
Seoul, South Korea;
Department of Business Administration, Soongsil University, Seoul, Korea
Simon Gim SNS Marketing Research Institute, Soongsil University, Seoul, Korea
Sunghwa Han Department of IT Policy and Management, Soongsil University,
Seoul, South Korea
Suyeon Han Department of Convergence Engneering, Hoseo Graduate School of
Venture, Seoul, Korea
Donghyuk Jo Department of Business Administration, Soongsil University,
Seoul, South Korea
Juyoung Kang Department of e-Business, Ajou University, Suwon, Republic of
Korea
Dooyeon Kim Department of Convergence IT, Konyang University,
Chungchungnam-Do, Republic of Korea
Yongmuk Kim Soongsil University, Seoul, South Korea

xi
xii Contributors

Hansol Ko Department of e-Business, Ajou University, Suwon, Republic of

Korea
Hanyoung Lee Department of Convergence Engineering, Hoseo Graduate School
of Venture, Seoul, South Korea
Kyoungho Lee Graduate School of Information Security, Korea University,
Seoul, Republic of Korea
Seulki Lee Department of Business Administration, Sangmyung University,
Seoul, Korea
Siyoung Lee Department of IT Policy and Management, Graduate School,
Soongsil University, Seoul, Korea
Donghyun Lim Department of Convergence Engineering, Hoseo Graduate School
of Venture, Seoul, Korea
Euntaek Lim Graduate School of Business Administration, Soongsil University,
Seoul, Korea
Jungkyu Moon Department of Convergence Engineering, Hoseo Graduate School
of Venture, Seoul, Korea
Kwansik Moon National Assembly, Seoul, Republic of Korea
Jungseon Oh KEPCO, ICT Planning, Seoul, Republic of Korea
Saeyong Oh Department of IT Policy and Management, Soongsil University,
Seoul, South Korea
Youngjoo Oh Department of Convergence Engineering, Hoseo Graduate School
of Venture, Seoul, Korea
Daewoo Park Department of Convergence Engineering, Hoseo Graduate School
of Venture, Seoul, Korea
Jongwoo Park Department of Business Administration, Soongsil University,
Seoul, South Korea
Hogun Rou Department of IT Policy and Management, Soongsil University,
Seoul, South Korea
Sungwoong Seo Hanwha Systems, Seoul, Korea
Hung-Trong Van Faculty of Digital Economy & E-Commerce, Vietnam–Korea
University of Information and Communication Technology, Danang, Vietnam
Thi-Thanh-Thao Vo Faculty of Digital Economy & E-Commerce, Vietnam–
Korea University of Information and Communication Technology, Danang,
Vietnam
Hyoboon Wang Soongsil University, Seoul, South Korea
Contributors xiii

Jinyoung Yang Soongsil University, Seoul, South Korea

Hyunjae Yoo Department of IT Policy and Management, Graduate School,
Soongsil University, Seoul, Korea
Hyunju Yoo Department of Convergence Engineering, Hoseo Graduate School of
Venture, Seoul, Korea
A Study on the Intention to Use Korean
Telemedicine Services: Focusing
on the UTAUT2 Model

Harim Byun and Jongwoo Park

Abstract With the outbreak of the recent COVID-19 pandemic, the importance
and demand of telemedicine, a method of supply medical services in a non-face-
to-face manner, is spreading worldwide. Despite its high medical standards and
IT technology, the telemedicine practice between medical staff and patients is not
allowed in Korea. Therefore, this study was intended to examine the factors influ-
encing on the intention to use Korean telemedicine services from the perspective
of potential consumers in China and Vietnam as a pavement basic work to expand
Korean telemedicine services overseas. To this end, the UTAUT2 model was recon-
structed to be suitable for the study on the acceptance of telemedicine service; users’
perceived perception on Korean telemedicine services (i.e. performance expectancy,
effort expectancy, social influence, price value, perceived risk) and personal charac-
teristics (i.e. innovativeness and concern for health) were set as main variables; and
their relationship with the use intention was empirically analyzed. As a result of the
analysis, it was found that, performance expectancy, social influence, price value, and
perceived risk factors, except for effort expectancy, have a significant relationship
influencing on the use intention. In addition, it was confirmed that innovativeness
and concern for health have a moderating effect on the relationship between perfor-
mance expectancy and use intention. Finally, specific implications derived based on
the analysis results are expected to be used as basic data for seeking global strate-
gies for Korean medical services in conjunction with the overseas expansion of the
Korean telemedicine service-related businesses.

Keywords COVID-19 · Telemedicine · UTAUT2 · Korean telemedicine service ·

Telehealth · Global medical service · Perceived risk · Use intention

H. Byun
Soongsil University, Seoul, South Korea
e-mail: [email protected]
J. Park (B)
Department of Business Administration, Soongsil University, Seoul, South Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 1

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_1
2 H. Byun and J. Park

1 Introduction

The recent outbreak of COVID-19 pandemic is accelerating a “zero contact society”

worldwide. Under the main word of ‘isolation and closure’, we are experiencing
changes in all lifestyles and industries (e.g. consumption, distribution, education,
etc.). In particular, as the necessity and market demand for the virtual telemedicine
in which medical staff and patients do not face each other have been confirmed, the
virtual telemedicine is rapid spreading worldwide, especially in the advanced [1].
Global Market Insights reported that the scale of the global digital healthcare
market is expected to grow from KRW130 trillion (USD16.4 Billion) in 2019 to
KRW600 trillion (USD5.404 Billion) in 2025, by an annual average of 29.6%; and
the scale of telemedicine market is also expected to reach $32.71 Billion by 2027
[2].
With its advantages, such as improving access to medical services and reducing
medical costs in addition to the prevention of infection on account of its non-contact
method, the telemedicine has emerged as a solution to the rapid increase in medical
expenses caused by the aging of the population, and the number of corporations and
governments in many countries promoting the telemedicine is increasing rapidly. In
particular, in countries with high-level medical and IT technologies (e.g. the United
States, Europe, and Japan), the telemedicine service is being promoted in response
to growing demand for efficient health care, and various business models are being
developed.
Korea has demonstrated its remarkable response capabilities based on rapid diag-
nosis, thorough quarantine and high-level medical standards during this COVID-
19, gaining credibility worldwide. As a result, K-bio and K-medicine of Korea
in conjunction with Korean Wave have gained international credibility. In addi-
tion, Korea, with high-level information and communication technology enough to
commercialize 5G for the first time in the world, has the best conditions to provide
telemedicine services beyond doubt. However, Korea is the only country in the world
that telemedicine practice between medical staff and patients is not allowed to date for
the reasons of the issues related to proving the effectiveness and safety of telemedicine
for many years [3]. In addition, on account of various legal regulations, the devel-
opment of related technologies, the scale of related corporation and investments and
the growth rate of corporation concerned are very low compared to other countries.
Now, the telemedicine service attracts intense attention as “must-have” rather than
“good-if-we-have” service before and after the outbreak of the global epidemic. In
order not fall behind such an inevitable global change, it is necessary to first develop
it as a global medical service business by turning our eyes to overseas markets where
telemedicine service can be more freely implemented, not to the domestic market. In
particular, telemedicine service can be intensely used as a strategic means to attract
overseas patients into Korea and to increase the possibility of the overseas expan-
sion of Korean medical institutions and medical consultations for potential medical
tourists are intensely being conducted remotely in preparation for the Post-COVID.
A Study on the Intention to Use Korean Telemedicine … 3

Studies on the acceptance and use of new information technologies are said to be
an advanced study field, but studies on the acceptance of technologies in the health
care have relatively seldom been conducted [4]. In particular, for the telemedicine,
studies have mainly focused on the possibility of technical implementation to date,
and studies on the inducement of practical use of technologies have been conducted
mostly on the Koreans for whom it is difficult to apply the telemedicine in Korea,
so they are considered to have limitations in light of the effectiveness of studies [5].
Therefore, this study has focused on China and Vietnam as main overseas market
for telemedicine service at the present time after thoroughly reviewing the rankings
of medical tourists visiting Korea by country, the status of the overseas expansion of
Korean medical institutions and the constraints on overseas telemedicine services (i.e.
accessibility, time difference between countries and the extent of telemedicine activa-
tion within the country), conducting an empirical on the potential customers in those
countries. In addition, this study intends to discover the acceptance factors related to
Korean telemedicine service and confirm the causal relationship with the use inten-
tion by using UTAUT (Unified Theory of Acceptance and Use of Technology). The
ultimate purpose of this study is to predict the possibility of the success of Korean
telemedicine service business in overseas markets; derive matters to supplement;
and ultimately provided basic data needed for seeking strategies to globalize Korean
medical services, including medical tourism.

2 Theoretical Background

2.1 The Concept and Definition of Telemedicine Service

There is no unified definition of the concept of telemedicine. In general, various

terminologies (i.e. Telemedicine, Telehealth, Telehealthcare, e-health and U-health)
are currently used interchangeably.
When classifying the concepts of the ‘telemedicine’ and the ‘telehealth’, which
are being most commonly used interchangeably, the telemedicine is a concept
that includes telehealth, and the telehealth occupies part of the telemedicine. The
telemedicine can be understood as the remote replacement of the treatment of patients
by doctors in an in-hospital clinic through various communication technologies (e.g.
video treatment, telephone treatment, remote secondary medical opinion and remote
prescription, etc.). On the other hand, the telemedicine is a comprehensive concept
that includes both remote patient monitoring and remote surgery, which are services
that remotely monitor patient health and disease in addition to treatment-related
services [3].
When looking at a few of conceptual definitions of telemedicine, Krupinski et al.
[6] defined the telemedicine as the “exchange of patient-related health information
between geographically distant medical service providers or between providers and
consumers (patients) through telecommunication technology and computers for the
4 H. Byun and J. Park

purpose of evaluation, diagnosis, treatment and education”, and the World Medical
Association (WMA) defined telemedicine as “a medical practice that determines and
recommends intervention, diagnosis and treatment for diseases based on clinical data,
records and other information transmitted from a distance through a telecommunica-
tion system”. Also the World Health Organization (WHO) defined the telemedicine as
the “act of all medical experts to prevent, diagnose, treat diseases or injuries; continu-
ously educate medical service providers; and provide useful information and medical
services for communities and local residents from distant places using information
and communications technologies” [7].
When looking at these definitions, the telemedicine is viewed differently, focusing
on the central functions performed (e.g. viewing the telemedicine as an alterna-
tive means of face-to-face treatment and viewing it as exchange of technology and
medical information), but what is common between various definitions is viewing the
telemedicine as the combination of medical service and information and communica-
tion technology. This study views the telemedicine from a comprehensive perspective
and defines the telemedicine as all medical-related activities performed in a non-
contact manner using information and communication technology between various
telemedicine service providers and telemedicine service recipients.

2.2 The Current Status of Overseas Telemedicine Service

Businesses at Home and Abroad

Thailand, Singapore, and India leading the global medical tourism market have intro-
duced the telemedicine from the beginning of 2000 as a countermeasure in response
to the gradual decline in the profits from medical tourists [8]. Table 1 briefly shows the

Table 1 Overseas telemedicine service status

Country (hospital) Contents
Thailand Through “Global Care Solution (GCS)” project (i.e. an online
(Bumrungrad Hospital) pre-treatment) promoted in collaboration with MS from Dec 2007, it
has strengthened its pre-treatment function, shortening waiting time
Singapore Through the operation of an online communication channel called
(Raffles Hospital) “ASK RMG” since 2008, medical staff responds in detail to inquiries
about health care from overseas customers in real time. This promotes
hospitals naturally, resulting in an increase in the number of recruited
customers
India Founding the “Remote Medical Network”, it has established
(Apollo Hospital) telemedicine service cooperation system in 9 countries (e.g.
Southwest/Central Asia, including Sri Lanka, Pakistan, Nepal,
Bangladesh, Myanmar, and Africa)
Source Taegyu et al. [8]
A Study on the Intention to Use Korean Telemedicine … 5

current status of the telemedicine services at the representative hospitals attracting

overseas patients in Thailand, Singapore and India.
In addition, Cuba is providing telemedicine services to countries in Central
America and the Caribbean, and Mexico has been conducting international
telemedicine services through cooperation with specific medical institutions in the
United States by intensively investing in the telemedicine infrastructure since 2014.
In the early 2000s, Korea’s overseas telemedicine service business began as part of
Official Development Assistance (ODA) project for countries with underdeveloped
health care environment, and as a result, the number of successful cases attracting
and treating patients in Korea has gradually increased [9].
As a result of the successful operation of the U-Health system, a remote video
treatment system installed in Vladivostok, Russia in 2011, Gangnam Severance
implements the personalized medical services for overseas patients, promoting the
recruitment of patients.
Seoul National University Hospital has been actively implementing remote
reading and remote consultations that receive medical opinions and consultation
from Seoul National University Hospital through a computational monitoring system
(PPCC: Pre-post Care Center) installed in the Sheikh Khalifa Hospital in the United
Arab Emirates, whose operation was commissioned to Seoul National University
Hospital from 2015. As above, Korean telemedicine service is generally promoted
by the connection between overseas governments and specific large domestic hospi-
tals under the policy intended to expand medical care for overseas patients. Large
domestic hospitals provide remote medical services similar to telemedicine services
by opening telemedicine centers through medical tourism centers and local hospitals
located overseas. Currently, domestic small and medium-sized hospitals (e.g. plastic
surgery, dermatology and oriental medicine clinics) mainly focus on remote consul-
tation using phone, e-mail, and social media for the purpose of attracting overseas
patients into Korea [10].

2.3 United Theory of the Acceptance and Use of Technology

The first theory intended to explain the phenomena of accepting new technologies
is the Technology Acceptance Model (TAM) suggested by Davis [11]. This model
was developed to apply the relationship between beliefs, attitudes, intentions and
actions to the acceptance of technology based on theory of rational behavior. The
TAM has been verified as a representative model with high explanatory power in
numerous technology acceptance studies since it was suggested. On the other hand,
controversy over the limitations of the proposed model in the organizational context
and the inability to sufficiently reflect the influence of various exogenous factors
has been raised steadily. As a supplement to this, Venkatesh et al. [12] suggested a
new technology acceptance model (UTAUT: Unified Theory of Acceptance and Use
of Technology) from an integrated perspective, but this model also focuses on the
factors influencing organization members’ acceptance of information technology.
6 H. Byun and J. Park

Since then, Venkatesh et al. [13] also presented the UTAUT2 (Extended Unified
Theory of Acceptance and Use of Technology) model that could better explain the
process of accepting and using technology by general consumers. In UTAUT2 model,
3 factors (i.e. hedonic motivation, price value and habit) were added to the 4 core
factors of the existing UTAUT model (i.e. performance expectancy, effort expectancy,
social influence and facilitating conditions) and these factors have been confirmed
to be critical variables in the results of recent researches using these factors [14–17].

2.4 Perceived Risks

The perceived risk initially suggested by Bauer [18] is the risk perceived subjectively
in the situation where consumers have multiple choices, referring to the uncertainty
and negative consequences that consumers feel when they cannot predict the outcome
of their actions. In other words, psychological discomfort and anxiety occurring as a
result of consumers’ perception of risk negatively influence the evaluation of product
or service, which may, in turn, directly influence the acceptance and purchase of
products and services [19]. Prior studies on the acceptance of technology applying
perceived risk suggested that the perceived risk reduces perceived usefulness and
intention to use and adopt, supporting the above argument [20–22]. The influence of
risk perception is more prominent in the situation of the purchase of services than in
tangible products and medical services are particularly high-risk services that make
it difficult to predict the quality of services until they are experienced [23]. The
telemedicine services using various information and communication technologies as
a medium are basically exposed to an open environment of so-called online, which
may make security main risk factors (e.g. hacking, errors, and personal (biometric)
information leakage). In addition, on account of the non-contact interactions between
medical staff and patients, the accuracy of communication, diagnosis, prescription
and resulting problem-solving issues may act as risk factors.

3 Research Model and Hypotheses

This study intends to investigate the factors influencing the intention of potential
overseas customers to use the telemedicine services to expand telemedicine services,
which are spreading as a new method of supplying medical services, into overseas
markets. Therefore, in this study, the UTAUT2 model was selected as a basic theo-
retical framework and new factors that could influence the use intention were added
to design a model for accepting Korean telemedicine services.
In this study, the 5 main factors set to influence the intention to use Korean
telemedicine services were ‘performance expectancy’, ‘effort expectancy’, ‘social
influence’, ‘price value’ and ‘perceived risk’. But the ‘hedonic motivation’ and ‘habit’
A Study on the Intention to Use Korean Telemedicine … 7

Antecedents of
Use Intention
Performance
Expectancy

Effort Expectancy
H1
Use Intention
Social Influence

Price Value

Perceived Risk H2 H3

Concern for Health Innovativeness

Fig. 1 Research model

factors in the UTAUT2 model were not reflected in this study because it was consid-
ered that the level of technological support expected at a personal level would not
greatly promote or delay consumers’ use of telemedicine on account of the service
supply path provided through medical institutions and telemedicine centers equipped
with many telemedicine systems, except for medical consultations that consumers
mainly use using mobile. In addition, ‘hedonic motivation’ and ‘habit’ were elim-
inated from the influencing factors in consideration of the essential characteristics
of medical treatment directly associated with human health and the situational char-
acteristics of the commercialization stage of telemedicine. As there are a number
of potential risk factors (e.g. threats to the health rights of patients and leakage
of personal (biometric) information due to remote medical services), which was
predicted to directly influence the user’s intention to use, the perceived risk factors
were added to the influencing factors for analysis. Figure 1 shows the research model
of this study.

4 Research Method

4.1 Data Collection and Analysis Method

In this study, general consumers in China and Vietnam were selected as potential
users of Korean telemedicine services, and a survey was conducted using a random
sampling method. The online survey was conducted for two weeks starting from
8 H. Byun and J. Park

Table 2 Sample characteristics

Item Sample size Ratio (%) Total
Country China 151 48.7 310
Vietnam 159 51.3
Gender Male 111 33.4 310
Female 199 59.9
Age 20’s 143 43.1 310
30’s 126 38.0
40’s 36 10.8
More than 50 5 1.5
Final education High school 32 9.6 310
College 58 17.5
University 149 44.9
Graduate school 71 21.4
Job Office work 79 23.8 310
Production occupation 17 5.1
Professional occupation 19 5.7
Student 105 31.6
Official 55 16.6
Self-employed 35 10.5

Jul 2, 2020, and finally 310 copies of data were used for analysis. In addition, 7-
point Likert scale was used to measure the responses to questions and SPSS 23.0
was used for statistical analysis. The demographic characteristics of the sample group
are summarized in Table 2.

4.2 Operational Definition of Variables

This study has selected ‘performance expectancy’, ‘effort expectancy’, ‘social influ-
ence’, ‘price value’ and ‘perceived risk’ as the main factors influencing the intention
of overseas consumers to use Korean telemedicine services through literature review
and listening to expert opinions; analyzed the relationship between the factors; and
additionally analyzed the moderating effect of personal characteristics (i.e. innova-
tiveness and concern for health) between these factors and the use intention. The
operational definitions of the main variables selected in this study are listed in
Table 3.
A Study on the Intention to Use Korean Telemedicine … 9

Table 3 Operational definition of variables

Variable Operational definition Source
Performance expectancy The extent to which the use of Korean telemedicine service [12, 24]
is believed to be able to help achieve the expected purpose
of use
Effort expectancy The extent to which Korean telemedicine services are [12, 25]
believed to be easy to use
Social influence The degree of belief that the reference group will consider [12, 26]
Korean telemedicine services positive and recommend to
their acquaintances the use of Korean telemedicine services
Price value The extent of satisfaction from using Korean telemedicine [14, 27]
services compared to the cost to pay for the use
Perceived risk The possibility of perceived loss from using Korean [28, 29]
telemedicine services
Innovativeness Individual’s disposition or willingness to quickly [20, 30]
experience new technologies and services
Concern for health The importance of health and the extent of concern for [31, 32]
health that individuals feel
Use intention The intention or plan to use Korean telemedicine services [11, 12]
now or in the future

5 Research Results

In this study, multiple regression analysis was conducted to test the hypotheses, and
the analysis results are listed in Tables 4 and 5. As a result of the analysis, it was
found that the performance expectancy, social influence, price value, and perceived
risk had a significant influence on the intention to use Korean telemedicine services.
In addition, as a result of analyzing the moderating effects of innovativeness and
concern for health, it was found that both innovativeness and concern for health had

Table 4 Results of hypotheses tests

Factor Use intention
B β t p
Performance expectancy 0.055 0.176 3.764** 0.000
Effort expectancy 0.054 0.031 0.571 0.569
Social influence 0.057 0.305 5.273** 0.000
Price value 0.057 0.344 6.094** 0.000
Perceived risk −0.039 −0.110 −2.959** 0.003
R2 0.650
F 112.418**
Note *p < 0.05, **p < 0.01
10 H. Byun and J. Park

Table 5 Results of moderating effect analysis

Path Moderator β t p
PE → UI Innovativeness 0.657 2.968** 0.003
EE → UI 0.073 0.323 0.747
SI → UI 0.101 0.444 0.657
PV → UI 0.001 0.003 0.998
PR → UI −0.332 −1.274 0.204
PE → UI Concern for health 0.572 2.466* 0.014
EE → UI 0.054 0.250 0.803
SI → UI −0.046 −0.213 0.831
PV → UI −0.099 −0.424 0.672
PR → UI −0.521 −1.873 0.062
Note *p < 0.05, **p < 0.01

a moderating effect on the relationship between performance expectancy and use

intention (t = 2.968, p = 0.003/t = 2.466, p = 014).

6 Conclusions

With the outbreak of COVID-19, the telemedicine is attracting more attention than
ever as it emerges as a safe and efficient method of supplying medical services. As a
basic study for the overseas expansion and promotion of telemedicine services that
are not allowed in Korea, this study was intended to investigate the factors influencing
the intention to use Korean telemedicine services from the perspective of potential
overseas customers. To this end, this study has tested the influence relationship by
reconstructing the UTAUT2 model suggested by Venkatesh et al. [13] to be suitable
for the studies on the acceptance of telemedicine services. The analysis results of
this study can be summarized as follow.
First, it was found that performance expectancy, social influence, price value, and
perceived risk had a significant influence on the intention to use Korean telemedicine
services. The significant influence relationship between performance expectancy and
intention to use has been proven in various studies on the acceptance of information
technology including the telemedicine [29, 33, 34]. Therefore, it deems necessary
to actively promote the usefulness of Korean telemedicine services and develop its
image based on the advices of medical experts and tests. In addition, management and
response strategies to remove or mitigate risk factors must be established within the
organization through the process of identifying risk factors perceived by consumers
in detail and predicting the consequences of risk according to the source.
A Study on the Intention to Use Korean Telemedicine … 11

Second, it was found that price value and social influence, in particular, greatly
influence the intention to use. Since these results indicate that the opinions of neigh-
bors and recommendation to them can be effectively used in the strategies for
promoting Korean telemedicine services, it is necessary to promote marketing strate-
gies that encourage sharing experience through Internet media and actively support
acquaintance recommendation programs. In addition, it was also found that overseas
consumers feel sensitive to the cost they will pay for using Korean telemedicine
services. This implies that establishing a rational cost strategy through meticulous
comparison of medical expenses in the local market and the telemedicine service
products provided in competing countries will become a key factor in expanding the
bases of overseas markets.
Third, it was found that the effort expectancy did not significantly influence the
intention to use Korean telemedicine services, being consistent with the results of
the study conducted by Choi et al. [16]. Given the fact that many of the survey
subjects are potential customers who have not actually used telemedicine services
and most of the telemedicine services, except for mobile medical consultations, are
provided at special institutions (e.g. medical institutions and telemedicine centers),
the difference in the extent of individual consumer’s effort to use the telemedicine
service is considered to be less relevant to the actual use.
Fourth, it was found that concern for health and innovativeness have a moder-
ating effect on the relationship between performance expectancy and intention to use
Korean telemedicine services. Accordingly, the need for an opportunity to directly
and indirectly experience Korean telemedicine through international fairs related to
beauty and health and promotion strategies for Korean telemedicine service products
that target users of healthcare-related devices and platforms is emphasized.
For the limitations of this study and future research directions are as follows. First,
since a comprehensive survey was conducted without distinguishing the nationality
(China, Vietnam) of the survey subjects, this study has limitations in generalizing the
study results. Second, this study has limitations in that it did not comprehensively
reflect various factors influencing the intention to use Korean telemedicine services,
an extended study to supplement these limitations will be required in the future.

References

1. Ahn S, Kim D, Kim B, Choi A (2021) US equity small cap inside 2021. YemoonBook
2. Stratistics Marker Research Consulting (2020) Telehealth-global market outlook (2018–2027)
3. Choi Y (2020) Digital health care: medical future. CloudNineBooks
4. Aggelidis VP, Prodromos DC (2009) Using a modified technology acceptance model in
hospitals. Int J Med Inform 78(2):115–126
5. Kim S (2010) Structural relationships among factor to adoption of telehealth service by health
conditions. Department of Public Health Graduate School, Inje University
6. Krupinski E, Nypaver R, Poropatich D, Ellis RS, Hasan S (2002) Clinical application in
telemedicine/telehealth. Telemedicine J e-Health 8(1):13–34
7. World Health Organization (2010) Telemedicine: opportunities and developments in member
states. Report on the second global survey on eHealth. Global observatory for eHealth series,
vol 2, Geneva, Switzerland
12 H. Byun and J. Park

8. Yu T, Choi Y (2014) Telemedicine centers as a venue for attracting international medical

tourists. Korean Public Manage Rev 28(3):133–167
9. Kim S (2017) Case study on problems and solutions of international telemedicine system in
Korean general hospitals. Graduate School of Public Administration Seoul National University
10. KukiNews (2020) Hospitals for remote counseling for overseas patients
11. Davis FD (1989) Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Q:319–340
12. Venkatesh V et al (2003) User acceptance of information technology: toward a unified view.
MIS Q 27(3):425–478
13. Venkatesh V, James YLT, Xin X (2012) Consumer acceptance and use of information
technology: extending the unified theory of acceptance and use of technology. MIS Q
36(1):157–178
14. Agarwal NK et al (2007) Factors affecting 3G adoption: an empirical study. PACIS 2007 Proc
3
15. Choi B, Kim H, Chung J (2019) Consumer acceptance of mobile gift certificates-focused on
UTAUT2. J Dig Converg 17(9):97–104
16. Choi W, Kim D, Choi S (2017) Understanding factors influencing usage and purchase intention
of a VR device: an extension of UTAUT2. J Korea Soc Comput Inf 18(3):173–208
17. Escobar-Rodríguez T, Elena CT (2013) Online drivers of consumer purchase of website airline
tickets. J Air Transp Manage 32:58–64
18. Bauer RA (1960) Consumer behavior as risk taking. In: Hancock R (ed) Proceedings of the
43rd America Marketing Association, pp 384–398
19. Featherman MS, Paul AP (2003) Predicting e-services adoption: a perceived risk facets
perspective. Int J Hum Comput Stud 59(4):451–474
20. Lu H, Chin-Lung H, Hsiu-Ying H (2005) An empirical study of the effect of perceived risk
upon intention to use online applications. Inf Manage Comput Secur 13(2):106–120
21. Crespo ÁH, Del Bosque IR, de los Salmones Sánchez MG (2009) The influence of perceived
risk on Internet shopping behavior: a multidimensional perspective. J Risk Rese 12(2):259–277
22. Kim H, Oh S (2019) Effects of gender difference and perceived risk on acceptance intention
of mobile easy payment service. J Manage Econ 41(3):145–165
23. Murray KB (1991) A test of services marketing theory: consumer information acquisition
activities. J Market 55(1):10–25
24. Kim S, Lee C (2013) Usage intention of u-healthcare service using unified theory of technology
adoption and usage. J Korea Contents Assoc 13(12):379–388
25. Rogers EM (2003) Diffusion of innovations, 5th edn. The Free Press, New York
26. Son H, Lee S, Cho M (2014) Influential factors of college students’ intention to use wearable
device-an application of the UTAUT2 model. Korean J Commun Inf 68:7–33
27. Khan IU, Zahid H, Safeer UK (2017) Understanding online banking adoption in a developing
country: UTAUT2 with cultural moderators. J Glob Inf Manage (JGIM) 25(1):43–65
28. Basoglu N, Daim TU, Topacan U (2012) Determining patient preferences for remote
monitoring. J Med Syst 36(3):1389–1401
29. Rho M (2013) A study of the expected benefit and perceived risk on telemedicine acceptance
for hypertension management. J Int Trade Commerce 9(1):343–361
30. Agarwal R, Jayesh P (1998) A conceptual and operational definition of personal innovativeness
in the domain of information technology. Inf Syst Res 9(2):204–215
31. Jayanti RK, Alvin CB (1998) The antecedents of preventive health care behavior: an empirical
study. J Acad Mark Sci 26(1):6–15
32. Shin M, Lee Y (2015) A study on the influential factors of purchase intention of wrist wearable
device. J Korea Contents Assoc 15(5):498–506
33. Chau PYK, Paul JH (2002) Examining a model of information technology acceptance by
individual professionals: an exploratory study. J Manage Inf Syst 18(4):191–229
34. Lanseng EJ, Tor WA (2007) Electronic healthcare: a study of people’s readiness and attitude
toward performing self-diagnosis. Int J Serv Indust Manage 18(4):394–417
A Structural Relationship Between
Environmental Uncertainty, Dynamic
Capability, and Business Performance
in a Smart Supply Chain Environment

Yongmuk Kim and Jongwoo Park

Abstract E-commerce, based in the 4th industrial revolution to respond to environ-

mental changes such as rapid technical development, increasing global competition,
and diverse customer needs, is accelerating while uncertainty in the business envi-
ronment grows more severe. In order to effectively respond to this uncertain business
environment, individual companies and parts of supply chains need flexibility, and
this need for dynamic flexibility is growing as uncertainty increases. In this study,
we tested our hypothesis using structural equation model analysis to verify the struc-
tural relationship between dynamic capability and business performance in the smart
supply chain environment of Korean small and medium manufacturers. We confirmed
through this that dynamic capability and environmental uncertainty in a smart supply
chain environment have a partially significant effect on business performance and
that manufacturers must strive to implement direct, systematic policies and improve
dynamic capability to respond to an uncertain business environment.

Keywords Smart supply chain · Dynamic capability · Environmental uncertainty ·

Business performance

1 Introduction

As industrial activity is paralyzed due to recent rapid, dynamic changes in the corpo-
rate environment, protectionism, epidemic status, etc., we have been facing crisis not
only in our daily lives but worldwide. In particular, untact, i.e. contactless, technology
is being promoted throughout society due to COVID-19. E-commerce based on the
4th industrial revolution is accelerating to respond to these changes, and uncertainty
in the business environment is growing. Studies proposing a clear definition of the 4th

Y. Kim
Soongsil University, Seoul, South Korea
e-mail: [email protected]
J. Park (B)
Department of Business Administration, Soongsil University, Seoul, South Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 13

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_2
14 Y. Kim and J. Park

industrial revolution in e-commerce have been few, but the 4th industrial revolution
features a high level of connectivity and superintelligence, meaning that everything is
connected by fusing digital devices, humanity, and physical environments to evolve
into an intellectualized society [1].
Despite this technical development, businesses’ supply chain structure has
comprised various participants from raw material producers to component producers,
finished item producers, wholesalers and retailers, and transport companies
throughout the process of manufacturing and delivering each product to the customer.
Various issues such as uncertainty in predicting demand, understock, overstock, and
logistics and transport delays may interfere with the operation of the supply chain
and cause uncertainty and complexity when its participants are not perfectly matched
[2–4]. The concept of smart supply chains has been proposed to reduce variations in
supply and demand occurring due to supply chain complexity and uncertainty [5].
Smart supply chains attach various sensors and RFID tags to devices and products
used in product manufacture and transport so relevant data can be collected and
analyzed to operate and make decisions based on meaningful data [5, 6]. It is impor-
tant for the technical capabilities of these smart supply chains to be able to respond
efficiently to insufficient resources. They can increase efficiency using smart tech-
nology, but only if the necessary internal core competences for decision-making and
for capabilities to respond to changes in the internal/external business environment
are in place first. Therefore, leading and participating companies in smart supply
chains must have dynamic capability in order to react flexibly to environmental
changes and achieve good performance [7].
Previous research on smart supply chains has focused on the use of digitalized
supply chains and studied, for example, the effects of information technology compe-
tence on supply chains; the effect of 4th industrial revolution technology on logistics
performance; the effect of IoT technology on supply chains; the effect of 4th indus-
trial revolution technology on supply chain performance; tracking logistics objects
for smart SCM; basic factors in smart supply chain composition; research on smart
SCM models using RFID; research on smart SCM models based on synchronization
of logistics information; and key technologies and supply chain management in the
4th industrial revolution [5, 7–13]. Empirical research to analyze structural relation-
ships between dynamic capability and business performance in a smart supply chain
environment is therefore lacking, and studies in this field are needed.
The purpose of this study is therefore to empirically analyze the effects of envi-
ronmental uncertainty and dynamic capability on business performance in a smart
supply chain environment. Through this study uncertain business environments and
proposes goals to achieve ongoing competitive advantages by confirming struc-
tural relationships connecting the acquisition of dynamic capability to business
performance.
A Structural Relationship Between Environmental … 15

2 Theoretical Background and Hypotheses

2.1 Smart Supply Chains

The huge wave of the 4th industrial revolution amplifying uncertainty as well as
changes in technology and markets has profound effects on companies of all sizes
[14]. As technology develops, intelligent supply chain management technology
is being used to strengthen competitiveness and ensure corporate productivity by
improving logistics information [8]. Smart supply chains are systems that fuse IoT
and cyber-physical systems with a flow of data and products with various sensors
or RFID tags attached to, for example, devices or products to solve problems of
dynamic complexity when operating the chains. Large volumes of information previ-
ously unavailable in supply chains are provided throughout the entire process so the
operator can use them to operate and make decisions. This reduces inefficiency and
enables flexible responses to business environments to achieve results [5]. Technolog-
ical development is accelerating changes to smart supply chains in order to respond
to uncertainty. If changes in large volumes of data generated between traders within
the smart supply chain are collected in real-time and used effectively, they can alle-
viate various issues such as supply–demand mismatch, overstock, understock, and
transport delays [5].

2.2 Dynamic Capabilities

Dynamic capabilities consist of information and knowledge acquired or learned to

control, integrate, or rearrange resources in order to achieve competitive advan-
tages while rapidly responding to a quickly changing business environment from an
extensive viewpoint of core competences, the foundation of resource-based theory.
They can be called systematic or organizational capabilities [15]. Dynamic capabili-
ties propose grounds for the need to strengthen and maintain capabilities and ensure
competitive advantages in very uncertain business environments while strengthening
the dynamics of needed internal capabilities [16]. The concept of dynamic capability
varies between scholars with a range of opinions regarding its constituents [17–21].
Exploring market opportunities is an essential element of sustainable corporate
growth, and if a company can locate an opening more quickly than its competitors, this
will be a good driver for its profit and growth [22]. The act of detecting market envi-
ronments and collecting data while pursuing growth is itself the process of exploring
opportunities [23]. Capability in exploring opportunities is a company’s action to
detect opportunities and risks in the changing market environment and appears as a
process of monitoring investment and technological changes to establish response
strategies and gain market information based on those market changes as well as
exchanging information about new opportunities [24]. From a smart supply chain
perspective, the disruptive information technology of the 4th industrial revolution is
16 Y. Kim and J. Park

created as a management system connected to the supply chain, and capability with
this smart IT affects the dynamic capability to be able to detect, explore, and seize
business opportunities and risk elements [16]. Therefore, in order for a company to
continue to thrive, it is important to constantly explore business opportunities, and
these capabilities may be accumulated through a focus on the process of pursuing
opportunities itself [25].
The construction and acquisition of resources and capabilities to respond to the
rapidly changing market environment and maintain ongoing competitive advantages
is an important factor for companies. Penrose [26] divided resources into physical
and human resources, and Wernerfelt [27] defined tangible and intangible assets
belonging semi-permanently to companies. The construction and acquisition of
resources and capabilities to respond to the rapidly changing market environment and
maintain ongoing competitive advantages is an important factor for companies. The
acquisition of external resources is an especially crucial factor for small and medium
businesses due to their relatively insufficient resources and capabilities [22]. Further-
more, acquisition of external resources and their application to the changing business
environment plays the role of leverage for the sake of a company’s survival, enabling
it to achieve better performance than its competitors. Resource acquisition capability,
the ability to acquire new information and knowledge from outside the company due
to insufficient resources in this changing environment, is a key factor in a company’s
dynamic capability [28].
For the sake of a company’s survival and sustainable growth in a quickly changing
market environment, reallocation of that company’s resources is inevitable [22].
This is because companies’ resources, capabilities, and structure must be integrated
or reconfigured from mature business or declining fields of business to new busi-
ness opportunities [24]. In other words, resource reconfiguration capabilities are a
company’s skill in integrating and reallocating previous and new resources to convert
into new resources [19, 24]. This conversion and adjustment of resources is necessary
for a company to survive and maintain its competitive advantages in a changing busi-
ness environment. A company’s skill in integrating and combining resources plays
a key role in product innovation. The process of product innovation demands active
combination of tangible and intangible assets by businesses and managers, and the
skill to combine these is at the heart of dynamic capability regardless of a company’s
size [29]. However, since small and medium companies are weaker than large ones
in terms of technology, capital, manpower, business, etc. [30], they may be limited
in their ability to transform according to smart supply chain environment changes
right away [31]. However, resource reconfiguration, a part of dynamic capability,
is an important factor in creating new technology through readjusting and reallo-
cating previously and newly acquired resources in order to respond to a changing
environment, even if such resources are limited [32].
A Structural Relationship Between Environmental … 17

2.3 Dynamic Capabilities and Business Performance

Dynamic capabilities are playing a larger role in business performance while levels
of uncertainty and competition increase in the business environment recently. Orga-
nizations must acquire dynamic capabilities for their own survival and performance
improvement in a rapidly changing business environment. Dynamic capabilities are
skills enabling a company to readjust its resources to achieve new competitive advan-
tages, and such capabilities are strategic rather than temporary problem-solving
abilities that show sustainability over time [18]. Companies must acquire enough
organizational capital to create excellent results while responding to environmental
changes and can only achieve sustained competitive advantages if they have the
dynamic capability to use this capital according to their vision and mission. Further-
more, dynamic capabilities are controlled by past activities, current position, and
future processes which are recreated in a new market environment [22]. Makadok
[33] stated that the capability to explore opportunities, resource acquisition, and the
capability to reconfigure resources, all elements of dynamic capability, show in a
company’s profits. Zahra et al. [34] proposed that dynamic capabilities indirectly
affect business performance. Kim et al. [22] stated that elements of dynamic capa-
bility have significant effects on a company’s financial and non-financial business
performance. It has additionally been proposed that a company’s dynamic capabil-
ities have a positive effect on its financial or non-financial performance and are an
element in its competitive advantages [35, 36]. This study therefore has the following
hypotheses based on previous research.
H1. Dynamic capabilities have a positive effect on financial performance.
H2. Dynamic capabilities have a positive effect on non-financial performance.

2.4 Environmental Uncertainty

The business environment is changing at such a rapid speed that it is difficult for
companies to predict by combining all elements related to direct/indirect business,
and the number of factors businesses must take into consideration is gradually
increasing as well [37]. Amidst this rapid change, environmental uncertainty from
the point of view of companies and the supply chain is an inherent condition in
interactions between companies, and this means that while new products and tech-
nologies are frequently released among fierce competition and customer demand
strongly fluctuates, government policies often change and the results of decisions
are difficult to accurately predict [38]. Ganbold and Matsui [39] subdivided envi-
ronmental uncertainty into supply uncertainty, demand uncertainty, and technical
uncertainty in a study on integration of supply chain management. In general, it
can be predicted that the more complex the environment and the more dynamic the
awareness, the more companies will strive to actively set targets and reduce that
uncertainty [40]. Hong and Cheong [41] proposed that the speed of environmental
18 Y. Kim and J. Park

change affects information sharing and cooperation, while Jo et al. [42] stated that
improvement of business skills is an essential factor to achieve goals in an uncer-
tain market environment and adapt to market changes. Furthermore, according to
research by Ji and Pyun [43], individual companies with resource limitations aim to
actively cope with environmental uncertainty to reduce it while accelerating product
release and decision-making and improving their responsivity to market changes in
order to use this environmental uncertainty as an opportunity. Lee [44] found that
the greater the uncertainty, the greater the logistics performance in a study of differ-
ences in logistics performance according to environmental uncertainty. Therefore,
the environmental uncertainty perceived by members of the supply chain can be
expected to affect dynamic capability and business performance, and so we propose
the following hypotheses.
H3. Environmental uncertainty has a regulative affect in the relationship between
dynamic capability and financial performance.
H4. Environmental uncertainty has a regulative affect in the relationship between
dynamic capability and non-financial performance.

3 Research Method

3.1 Data Collection

In A total of 290 surveys of employees of small and medium manufacturers under-

standing supply chains were distributed by mail from January to February of 2020 in
order to analyze the structural relationship between dynamic capability and business
performance in a smart supply chain environment. Of these, 219 were collected, and
8 were excluded due to unfaithful responses, leaving 211.
The manufacturers who were subjects of this study could be broadly classified
into suppliers and purchasers, with 83 suppliers (39.3%) and 128 purchasers (60.7%).
There were 159 companies less than 20 years old (75.4%) and 52 more than 20 years
old (24.6%). In terms of number of employees, 205 companies had less than 100
(97.2%), while 6 (0.8%) had more than 100. There were 149 companies (70.6%)
with less than 10 billion won in sales and 62 (29.4%) with more than 10 billion won
in sales. Regarding partners, 180 companies (85.3%) had less than 50 partners, while
31 companies (14.7%) had more than 50 partners.

3.2 Measurement

The variables of dynamic capability and business performance used in this study
were derived from previous research and measured on a 5-point Likert scale. Items
A Structural Relationship Between Environmental … 19

measured for each variable were tested for validity and reliability in preceding
research, and these were adjusted and used for this study (Table 1).

4 Empirical Analysis

4.1 Testing the Measurement Model

To test this study’s hypotheses, we had to test the fitness, reliability, and validity
of the measurement model, and we used confirmatory factor analysis for this. We
judged fitness of the measurement model by using absolute fit index, incremental fit
index, and parsimonious fit index. The results of confirmatory factor analysis of the
measurement model showed that χ2 = 1709.131 (P = 0.000) and χ2/df = 2.735,
below the recommended level of 3, and RMR = 0.054, below the recommended
figure of 0.08. Furthermore, PGFI = 0.607, above the recommended value of 0.6;
CFI = 0.849 and TLI = 0.830, slightly lower than the recommended figure of 0.9
but still satisfactory; and PNFI = 0.697 and PCFI = 0.755, above the recommended
level of 0.6. The measurement model proposed in this study was thus deemed to be
fit. The reliability and validity of the measurement tool were thereby proven as the
measurement model was deemed fit. The convergent validity of the measurement tool
must have a standardized regression coefficient greater than 0.5 and ideally greater
than 0.7. Construct reliability (CR) must be above 0.7 and average variance extracted
(AVE) must be above 0.5 to be considered fit. Furthermore, for discriminant validity,
the latent factor’s AVE must exceed the squared value of the correlation coefficient
between constructs (∅2).

4.2 Testing the Structural Model

With the fitness of this study’s measurement model and reliability and validity of
this model confirmed, we implemented analysis of the structural model to clarify
whether the theoretical relationships of the research model were supported by the
data. The results of testing the fitness of the structural model showed that χ2 =
1703.81 (P = 0.000) and χ2/df = 2.734, below the recommended level of 3, and
RMR = 0.056, below the recommended figure of 0.08. Furthermore, PGFI = 0.612,
above the recommended value of 0.6, and CFI = 0.847 and TLI = 0.830, slightly
below the recommended level of 0.9 but still satisfactory. PNFI = 0.703 and PCFI
= 0.763, above the recommended value of 0.6, meaning that the proposed structural
model is fit.
With the structural model used in this study deemed fit, we tested our hypotheses
by measuring path coefficients between the research variables. Results of testing the
hypotheses showed that in the relationship between dynamic capability and financial
20 Y. Kim and J. Park

Table 1 Sample characteristics

Variable Item measured References
Dynamic capability Opportunity exploration Benchmarking excellence in [17, 18, 45, 24, 46, 22]
capability the same industry
Occasional checking and
analysis of market changes
Seeking ideas for new
product development
Ongoing collection of
competitor information
Exploring, collecting, and
analyzing market
opportunities
Resource acquisition Fostering experts for new
capability business
Acquiring external knowledge
for new product development
Building partnerships with
other companies to develop
capabilities
Systematizing internal
knowledge from employee
experiences
Acquiring resources for new
product development
Resource reconfiguration Process reconfiguration to
capability introduce new technology and
knowledge
Integration and adjusting of
management teams to respond
to environmental changes
Resource reallocation to suit
new environments
Changing technical facility
processes to respond to
environmental changes
Reconfiguration of resources
needed to develop new
products
Business performance Financial performance Increased sales [22, 47, 48]
Increased operating profit
Improved cash flow
Increased market share
Non-financial Improving participants’
performance capabilities and technical
level
Improving customer
satisfaction
Improving the company and
product image
Improving employee
satisfaction
Environmental uncertainty Demand uncertainty [2, 4, 49]
Supply uncertainty
Competitor uncertainty
Technical uncertainty
A Structural Relationship Between Environmental … 21

performance, the coefficient value of opportunity exploration capabilities toward

financial performance was (β) = 0.349 (C.R. = 3.016, p = 0.003), the coefficient
value of resource acquisition capability toward financial performance was (β) =
0.064 (C.R. = 0.454, p = 0.650), and the coefficient value of resource reconfiguration
capability toward financial performance was (β) = 0.223 (C.R. = 1.940, p = 0.052),
and we therefore selected Hypothesis H1-1 and rejected Hypotheses H1-2 and H1-3.
Moreover, in the relationship between dynamic capability and non-financial
performance, the coefficient value of opportunity exploration capability toward non-
financial performance was (β) = 0.329 (C.R. = 3.535, p = 0.000), the coefficient
value of resource acquisition capability toward non-financial performance was (β)
= 0.265 (C.R. = 2.303, p = 0.021), and the coefficient value of resource reconfigu-
ration capability toward non-financial performance was (β) = 0.276 (C.R. = 2.966,
p = 0.003), and we therefore selected all three Hypotheses H2-1, H2-2, and H2-3.

4.3 Moderating Effects of Environmental Uncertainty

Results of analyzing the moderating effects of environmental uncertainty in the rela-

tionship between dynamic capability and business performance showed a correlation
between opportunity exploration capability and non-financial performance (t = −
2.026, p = 0.044) and a correlation between resource reconfiguration capability and
non-financial performance (t = −2.850, p = 0.005), suggesting that environmental
uncertainty has a significant regulating effect.

5 Conclusions

For this study, we reviewed the role of variables through literature research and
empirically tested their influence relationships to determine the effect of a company’s
dynamic capabilities on its business performance so small and medium manufacturers
can respond to rapidly changing business environments within a smart supply chain.
Setting hypotheses regarding the interaction between dynamic capability and busi-
ness performance showed that dynamic capabilities had a partially significant effect
on business performance. In more detail, opportunity exploration capabilities affected
both financial and non-financial performance, and resource acquisition capabilities
and resource reconfiguration capabilities only affected non-financial performance.
These results support the conclusions of Eisenhardt and Martin [18], Kim and Huh
[36], and Kim et al. [22] and confirm that small and medium companies’ dynamic
capabilities are positively correlated with their business performance. As small and
medium companies are relatively lacking in resources and struggle to sustain compet-
itive advantages in a rapidly changing, dynamic environment, they need dynamic
capabilities to respond to these constantly changing situations. However, resource
acquisition capabilities and resource reconfiguration capabilities were not found to
22 Y. Kim and J. Park

have a significant effect on financial performance. We can infer from this that since
financial performance is a short-term, direct measure of a company’s past business
activities, there is little direct, short-term correlation between resource acquisition
capabilities and resource reconfiguration capabilities which appropriately reallocate
to a company’s business capabilities and act as leverage.
Moreover, the results of testing the regulating effects of environmental uncertainty
in the relationship between dynamic capabilities and business performance showed
that environmental uncertainty does have a regulating effect only in the relationship
between dynamic capability and non-financial performance. This shows that environ-
mental uncertainty has a regulating effect in the more long-term results appearing in
the relationship with non-financial performance rather than in the short-term results
of financial performance, and this could imply a need to establish strategies early in
order to improve long-term responsiveness to environmental uncertainty.
This study deviated from previous research which focused on the role of IT on
supply chain performance and logistics performance in a smart supply chain envi-
ronment for the sake of small and medium companies’ business performance to
confirm dynamic capability factors that companies must improve to maintain compet-
itive advantages in an uncertain business environment and to establish a structural
relationship linking acquisition of dynamic capabilities to business performance.
For this study, we set our primary subjects as small and medium manufacturers to
establish a structural relationship between dynamic capabilities and business perfor-
mance in a smart supply chain environment. However, the business conditions of
these companies may differ, and we did not measure or categorize them systemati-
cally according to their stage of smart supply chain construction but rather general-
ized them all. More meaningful studies are expected in the future if differences in
awareness of dynamic capabilities is confirmed by classifying clear differences in
the characteristics of different industries and the roles of suppliers and purchasers as
we selected suppliers and purchasers in the supply chain environment based on their
subjective responses.

References

1. Kim JH (2016) In the era of the 4th industrial revolution, seeking strategic responses to future
social changes. R&D InI 15:45–58
2. Kim SO, Youn SH (2008) A study on the effects of environment uncertainty and inter-
firm collaboration practice on supply chain flexibility. Korean Assoc Indust Bus Administ
23(1):337–364
3. Lee GD (2013) An empirical study on the relationships among environmental uncertainty,
management by objectives and organization satisfaction. Tax Account Res 22(1):179–202
4. Kim JY, Bang HY (2014) The effects of output sector uncertainty on dependence, commitment
and strategic performance: a comparative analysis on korean and american manufacturers.
Korea Assoc Int Commer Inf 16(1):163–183
5. Shin JC, Lim OK, Park YH, Song SH (2017) A study on determining priorities of basic factors
for implementing smart supply chain. J Korean Soc Supply Chain Manage 17(1):1–12
A Structural Relationship Between Environmental … 23

6. Lifang W, Yue X, Alan J, Yen DC (2016) Smart supply chain management: a review and
implications for future research. Int J Logist Manage 27(2):395–417
7. Moon TS, Kang SB (2014) An empirical study on the impact of it competence on supply chain
performance through supply chain dynamic capabilities. Korean Manage Rev 43(1):245–272
8. Kim JG (2013) A study on the logistics information synchronization based smart SCM model.
KIPS Trans Softw Data Eng:311–318
9. Lee KB, Baek DH, Kim DH (2016) A study on the effect of the IoT technology on SCM. J Inf
Technol Serv 15(1):227–243
10. Lee CB, Noh JH, Kim JH (2017) A study on the perception of the impact of fourth industrial
revolution on the performance of logistics management. Korea Logist Rev 27(5):1–12
11. Chung CC (2017) The fourth industrial revolution—an exploratory study on main technologies
and supply chain management. Korea Logist Rev 27(6):193–209
12. Noh JH, Lee KN (2018) The effect of supply chain related 4th industrial revolution technology
on BSC supply chain performance. Korea Logist Rev 28(5):53–64
13. Kwak KG, Hwang SY, Shin DJ, Park KW, Kim JJ, Park JM (2020) Study of logistics object
tracking service for smart SCM. J Korean Inst Indust Eng 46(1):78–81
14. Kim HT, Kwon SJ (2019) Exploring the industrial structure and innovation policy for
sustainable growth of machinery industry in South Korea. Sci Technol Policy Inst 2(1):107–131
15. Huh YH, Lee C (2012) Determinants of dynamic capability and its relationships with
competitive advantage and performance in foreign markets. Int Bus J 23(1):33–73
16. Rhee YP (2020) The relationships between IT capability, dynamic capability and international
performance in Korean SMEs. Int Bus Rev 24(1):107–120
17. Teece DJ, Pisano G, Shuen A (1997) Dynamic capabilities and strategic management. Strateg
Manage J 18(7):509–533
18. Eisenhardt KM, Martin JA (2000) Dynamic capabilities: what are they? Strateg Manage J
21(10):1105–1121
19. Bowman C, Ambrosini V (2003) How the resource-based and the dynamic capability views of
the firm inform corporate-level strategy. Br J Manage 14(4):289–303
20. Wang C, Ahmed PK (2007) Dynamic capabilities: a review and research agenda. Int J Manage
Rev 9(1):31–51
21. Denford JS (2013) Building knowledge: developing a knowledge-based dynamic capabilities
typology. J Knowl Manage 17(2):175–194
22. Kim JK, Yang HC, Ahn TD (2017) The moderating effect of manufacturing process type on
the relationship of dynamic capabilities and business performance of SMEs. Asia-Pacific J
Multimedia Serv Converg Art Human Sociol 7(8):141–151
23. Helfat C, Finkelstein S, Mitchell W, Peteraf M, Singh H, Teece D, Winter S (2007) Dynamic
capabilities: understanding strategic change in organizations, pp 1–18
24. Teece DJ (2007) Explicating dynamic capabilities: the nature and microfoundations of
(sustainable) enterprise performance. Strateg Manage J 28(13):1319–1350
25. Ambrosini V, Bowman C (2009) What are dynamic capabilities and are they a useful construct
in strategic management? Int J Manage Rev 11(1):29–49
26. Penrose ET (1959) The theory of the growth of the firm. Oxford
27. Wernerfelt B (1984) A resource-based view of the firm. Strateg Manage J 5(2):171–180
28. Verona G, Ravasi D (2003) Unbundling dynamic capabilities: an exploratory study of
continuous product innovation. Indust Corpor Change 12(3):577–606
29. Augier M, Teece DJ (2009) Dynamic capabilities and the role of managers in business strategy
and economic performance. Organ Sci 20(2):410–421
30. Kim YY, Park YS (2017) Fourth industrial revolution and SME supporting policy 20(2):387–
405
31. Park CK, Kim CB (2019) Effect of agility capabilities for smart manufacturing and smart SCM
implication on enterprise and supply chain performance in the 4th industrial revolution. J Small
Bus Innov 22(4):23–67
32. Morgan NA (2012) Marketing and business performance. J Acad Mark Sci 40(1):102–119
24 Y. Kim and J. Park

33. Makadok R (2001) Toward a synthesis of the resource-based and dynamic capability views of
rent creation. Strateg Manage J 22:387–401
34. Zahra SA, Sapienza HJ, Davidsson P (2006) Entrepreneurship and dynamic capabilities: a
review, model and research agenda. J Manage Stud 43(4):917–955
35. Hwang KY, Sung EH (2015) The relationships between dynamic capabilities, innovation
performance and performance of export venture firms. J Int Trade Commer 11:401–420
36. Kim GT, Huh MG (2016) Dynamic capabilities and competitive advantages: the moderating
effect of environmental dynamism. J Strat Manage 19(3):81–103
37. Huh SJ, Lee JY, Hyeon JW, Choi YS (2018) Business environment uncertainty and real earnings
management. Korean Corpor Manage Rev 25(6):77–101
38. Miller D (1988) Relating Porter’s business strategies to environment and structure: analysis
ans performance implications. Acad Manage J 31(2):280–308
39. Ganbold O, Matsui Y (2017) Impact of environmental uncertainty on supply chain integration:
empirical evidence. J Jpns Oper Manage Strat 7(1):37–56
40. Park SM, Park CW (2018) A study on the effect of multi-dimensionality of environmental
uncertainty on smart SCM factors and corporate management performance. Korea J Logist
26(2):105–126
41. Hong KS, Cheong KW (2004) The impact of environmental clockspeed, information sharing,
and collaboration on supply chain performance. Korean Small Bus Rev 26(2):77–100
42. Jo YG, Lee HG, Ha KT (2007) The relationship among the utilization of IT based on realizing
RTE, agility capabilities and company performance. Entrue J Inf Technol 6(2):113–127
43. Ji SG, Pyun HS (2009) The relationship among environmental uncertainty, marketing agility,
marketing performance. Korea J Bus Administ 22(2):1013–1035
44. Lee CS (2012) The moderating effects of logistics management system sophistication and
strategy on the relationship of environmental uncertainty and logistics performance. Korea
Logist Rev 22(1):183–209
45. Zollo M, Winter S (2002) Deliberate learning and the evolution of dynamic capabilities. Organ
Sci 13:339–351
46. Chun JI, Lee BH (2016) The effect of dynamic capabilities on international performance of
Korean exporting SMEs: the moderating role of environmental dynamism and firm type. Int
Bus Rev 20(1):45–74
47. Karplan RS, Norton DP (1966) Using the balanced scorecard as a strategic management system.
Harvard Bus Rev:75–85
48. Lee JD, Lee YB, Bae YS (2014) The effects of SMEs’ core competency and competitive
strategy on their business performance. J Korean Entrepreneurs Soc 9(3):154–183
49. Walker G, Weber D (1984) A transaction cost approach to make-or-buy decision. Adm Sci Q
29(3):374–391
Analysis of the Relative Importance
of Coaching Service Quality

Jinyoung Yang, Hyoboon Wang, and Donghyuk Jo

Abstract This study aims to analyze the relative importance between determinants
and factors of coaching service quality. For this purpose, the hierarchical factors of
coaching service quality are drawn by reviewing literature on coaching and using
Delphi technique, and survey was conducted on expert coaches in the coaching field to
analyze the relative importance of coaching service quality. As a result of analysis,
the relative important of primary hierarchical factors of coaching service quality
factors were Reliability, Expertise, Empathy, Responsive-ness, Effective-ness, and
Tangibles from highest to lowest. In addition, the result of combined importance on
secondary hierarchical factors showed Listening Actively, Positive Feedback, and
Ethical Practice as important factors. Such results may be interpreted as establishing
foundation of coaching relationship and forming close confidential relationships are
important in improving coaching effects. Result of this study may be used as basic
data in establishing an academic knowledge system of coaching and expanding the
base for coaching.

Keywords Coaching · Coaching competency · Coaching service quality · AHP

1 Introduction

The modern era, which began at the end of the nineteenth century, brought upon
signifi-cant socioeconomic changes unlike the previous society due to the effects
of logical-rational thinking, and scientific positivism, and the world following it

J. Yang · H. Wang
Soongsil University, Seoul, South Korea
e-mail: [email protected]
H. Wang
e-mail: [email protected]
D. Jo (B)
Department of Business Administration, Soongsil University, Seoul, South Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 25

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_3
26 J. Yang et al.

is characterized as two significant changes as the beginning of postmodern era in

the mid-twentieth century, which one of them is the transformation from industrial
economy to service-based economy, and another one is the appearance of coaching
industry in more complicated and growing social changes. Whereas the focus of
economic activity in the modern era was the production of goods and money, it shifted
to the production of nonmaterial convenience based on knowledge, information, and
technology in the postmodern era. Also, people became aware through World War I
and II that the rationality and reason of humans in the existing era don’t just bring
positive results, and unlike the modern era where an absolute standard or truth existed
and laborers were considered production unit under hierarchical order and control,
postmodern era formed a service-oriented society by acknowledging individual’s
diversity and changing the nature and idea of labor through reintegration of life and
work [1–3].
Second important change in postmodern era is the appearance of coaching. Soci-
oeconomic change in this era change the style and pattern of life, and brought signif-
icant changes in new perspective and value. Individual value became important in
complex globalization, and human’s consciousness gradually grew as the quality of
life improved due to the increase of economic wealth. Growth of human’s conscious-
ness lead to the desire for respect and recognition beyond survival, and to the interest
in social success, and people began to pay attention to self-improvement and growth
as the meaning of life, happiness and desire for self-realization grew larger [4]. As a
result, coaching focusing on growth and development came to the fore, and is being
used in different areas including management, life, career, business, Christianity,
relationship, etc. [5].
Service is a tangible and intangible substance that satisfies consumer’s desire
for expectation, and may be understood as a series efforts to satisfy clients [6].
It also evaluates and manages service quality in various service areas in order to
satisfy client’s expectation and draw satisfaction. Studies related to service quality
is particularly active in medicine, and hospital management competitiveness is in-
creased by acquiring positive effects in client loyalty and word of mouth effects
through improved reuse intention beyond client satisfaction by accurately under-
standing, improving, and developing client’s demand and necessity [7]. On the other
hand, in coaching, objective evaluations are not made how the expectation of clients
who received coaching service is satisfied. Mostly, client feedback is received on
coaching service after agreed coaching sessions, but they mostly consist of subjective
evaluations such as what they liked or self-insight, so there are still insufficient
detailed and objective data on the effects of coaching and client satisfaction.
Therefore, this study intends to establish and develop service quality of coaching
from service perspective based on coaching competency, and analyze the relative
importance of coaching service quality using AHP analysis. The purpose of this
pro-cess is to contribute to expanding base for coaching through the improvement
of coaching quality by measuring and managing service quality, positive effects
on client loyalty beyond client satisfaction, and establishment of quality factors of
coaching service.
Analysis of the Relative Importance of Coaching Service Quality 27

2 Theoretical Background

2.1 Coaching

The term coaching originated from a Hungarian four-wheeled carriage in the

sixteenth century called kocsi, and refers to a personalized transportation service
that departs from passenger’s current position and arrives at the desired destination
[8]. Since coaching is customized to coach, passenger, and situation, the definition
of coaching has been suggested differently by scholars of different areas based on
its nature and object, combined with academic ground theory. Based on various
perspectives of scholars, definition of coaching can be categorized into the aspect
that emphasizes promotion of study, aspect that emphasizes leadership, and aspect
that emphasizes communication [9].
First is the definition of coaching that supports and promotes learning [10]. In this
aspect, coaching is not teaching, but awakening of individual’s potentials to help one
study by oneself. In other words, it is a system that supports client’s self-realization
through a systematic and cooperative process based on problem-solving through indi-
vidual’s growth and self-directed learning experience [11]. Second is the definition
of coaching approaching as a leadership type. In this perspective, coaching promotes
client’s behavioral change through conversation and discussion, to empower for
further achievement [12]. Last is the definition of coaching as a communication
skill. In this approach, coaching is a two-way communication skill based on respect
for client, paying attention client’s desire and promoting client’s voluntary behavior,
and is a communication process that enables continuous development of competence
[13].
In addition, International Coach Federation (ICF) defines coaching as a profes-
sional relationship which helps achieve extraordinary results in life, career, business,
or organization, and Korea Coach Association (KCA) defines coaching as a hori-
zontal partnership that helps realize the highest values by maximizing individual’s
and organization’s potentials [14, 15]. Like this, definitions of coaching varies based
on perspective, but the common key elements of coaching may be summarized as
a partnership that supports an individual and an organization to demonstrate one’s
potentials to grow and develop, and a two-way communication system [16].
Coaching is different from similar areas such as counseling, education, training,
consulting, and mentoring in the fact that it offers a horizontal partnership, two-way
communication, client as the subject of problem-solving, growth and development,
client’s voluntary and proactive participation, minimal intervention of an expert, and
active delegation of authority to a client [17].
28 J. Yang et al.

2.2 Coaching Competency

Generally, competency is divided into an observable overt element that can success-
fully perform a certain task and an unobservable inherent characteristic [18], and
coaching competency refers to the intrinsic and extrinsic characteristics of a coach
including values, attitude, knowledge, and techniques, as a whole [19].
In November 2019, ICF categorized core competencies of a professional coach
into 8 characteristics. The 8 core competencies include demonstrating ethical prac-
tice, embodying a coaching mindset, establishing and maintaining agreements, culti-
vating trust and safety, maintaining presence, listening actively, evoking awareness,
and facilitating client growth, which are supported with 63 sub-guidelines to provide
the standard of competencies as a professional coach [14]. In Korea, KCA devel-
oped a standard coaching competency model for Korean coaches in 2019 through
experts’ validity verification. It divides the competency into ‘Being a coach’ and
‘Coaching’, where ‘Being a coach’ consists of the ability to ‘realize philosophy of
coaching’, coach’s ‘self-management’ ability, ability to ‘detect’ referring to coach’s
awareness and re-flection using intuition, and coach’s ‘expertise’ referring to accu-
mulation and use of knowledge and experience for coaching operation. ‘Coaching’
competency consists of ‘participation’ ability referring to a horizontal partnership
participating in client’s changes and growth, ‘expansion of consciousness’ ability to
help client’s change in perspective and expansion of awareness, ‘communication’
ability enhancing coaching effects such as effective questions and active listening,
and ‘process management’ ability to help client achieve goal [15].
Stowell [12] considered establishment of partnership, cooperation, interest in
demand, empathy, compliment, patience, and creation of environment for interaction
as important factors of coaching competency, and Law [20] considered empathy, kind
personality, patience for ambiguity, awareness, diplomatic ability, facilitation skill,
learning skill, interview skill, listening skill, and change management as factors of
coaching competency. Song [21] coaching ethics, listening skill, questioning skill,
acknowledging skill, feedback skill, summarizing skill, process application skill,
skill to use coaching diagnostic tool, time management skill, business insight, under-
standing of organizational issue, strategic thinking, coaching performance mainte-
nance and management, and ability use experience as important competencies. Do
and Kim [22] deduced autonomy, optimism, hope, relationship, and listening to
question and feedback as components of coaching competency.

2.3 Coaching Service Quality

As industrial economy transformed into service-based economy, the importance of

service began to rise, and the service society, formed as a result of huge socioe-
conomic changes, shifted from provider-centered to consumer-centered and grew
to-wards client-oriented perspective.
Analysis of the Relative Importance of Coaching Service Quality 29

Service quality emphasizes subjective judgment based on client’s individuality

and diversity and is difficult to measure objectively, therefore, it needs to be evalu-
ated through indirect clues. In other words, it is measured by subjective expectation
and perception of a client who experienced the service. In terms of various previous
studies on service quality, Grönroos [23] called it the result of comparison evalua-
tion process client’s perceived service and expected service, Parasuraman et al. [24]
explained service quality as the degree and direction of difference between expected
service and perceived service, and defined service quality as client’s overall judgment
or attitude towards the excellence of service. Cronin and Talor [25] raised an objec-
tion to Parasuraman et al. [24] expectation and performance management before
and after client’s service experience, and insisted that it needs to be measured only
through the perceived service performance, eliminating the expectation on service.
After, studies have been conducted in the 2000s, and Shen et al. [26] defined it as the
degree of satisfaction beyond client’s expectation, and stated that service quality can
be measured through client satisfaction. Study by Etgar and Fuchs [27] explained
that it connotes psychological decision on service perceived by client’s expectation.
As interest and researches on service quality became active, attempts to establish
and measure quality factors have been made. The model most frequently used in
studies is SERVQUAL model by Parasuraman et al. Their early study suggested
10 quality factors perceived by clients who experienced service, but were criti-
cized that service expectation performance measurement based on experience may
be difficult to apply in actual field, and revised them into 5 factors in the follow-
up study, including tangibles, reliability, responsiveness, assurance, and empathy.
SERVQUAL model is usefully utilized to measure and manage service quality in
various service areas, and contributes significantly to management performance by
studying client’s desire, need, and expectations, improve client satisfaction, and se-
cure loyal clients; it is widely used in studies of different areas from management to
hotel, education, medicine, beauty, and counseling.
Coaching is a horizontal conversation with a process, which is provided to help
achieve client’s desired goal and future-oriented growth. Clients receiving coaching
can establish detailed implementation plans for goals and realize voluntary actions
by introspect themselves again and discover possibility and potential in themselves.
Performance of coaching may be evaluated based on ‘how much client changed
or grew’, and coaching may be considered successful when the client who received
coaching seeks a new life to achieve the goal through different thinking and behavior.
For successful coaching, coaches must possess coaching competencies to interact
with clients and implement in the coaching scene, and success of coaching can be
expected when they properly possess and realize such competencies. Eventually,
the quality of coaching needs to be improved in order to bring benefits and satis-
faction to the client and improve coaching performance. In other words, coaching
competency can improve the quality of coaching service and enhance customer satis-
faction, therefore, coaching competency-based service quality needs to be developed,
and coaching as an area of service industry needs studies that establish, measure, and
evaluate quality factors of coaching service using the service quality measurement
model.
30 J. Yang et al.

1 Step 2 Step 3 Step

Literature review, used Delphi technique Used AHP teiques

research of previous study identifying secondary hierar- measurement the
data, drawing primary chical factors of coaching relative importance
hierarchical factors of coach- service quality from between identified
ing service quality expert groups hierarchies.

Fig. 1 Coaching service quality research methods and procedures

3 Research Design

3.1 Research Method

To evaluate the quality factors of coaching service and relative important, the study
was designed and implemented in 3 steps. First step includes literature review and
research of previous study data, and analyzing core coaching competencies by ICF
and coaching competencies by KCA to draw primary hierarchical factors of coaching
competency-based coaching service quality. Second step used Delphi technique.
Delphi survey was collected through 2 professors of business administration and
13 coaching experts with doctorate related to coaching to draw secondary hier-
archical factors of coaching service quality. Third step used AHP techniques to
analyze the relative importance between hierarchies. Based on final hierarchy struc-
ture deduced through Delphi technique, a 9-point scale survey was written on the
relative importance of evaluation factors related to coaching service quality.
AHP technique was developed by Thomas L. Satty, and is still widely used and
studied in decision-making area; the biggest strength of AHP technique is the reflec-
tion of qualitative and quantitative evaluations on various alternatives. Service quality
for clients is intangible and contains psychological nature, and using AHP technique
can effectively analyze the psychological area that reflects individual’s subjective
thinking, emotion, and belief (Fig. 1).

3.2 Study Model

For this study, the SERVQUAL model, data from previous studies, opinions from
coaching experts, and Delphi questionnaire were comprehensively reflected and
developed as shown in Table 1.
Analysis of the Relative Importance of Coaching Service Quality 31

Table 1 Hierarchical structure of coaching service quality

Coaching service quality Primary hierarchical factors Secondary hierarchical factors
Expertise Coaching knowledge and
experience
Listening actively
Effective question
Intuition and Insight
Tangibles Coach’s attire and appearance
Proper place and atmosphere for
coaching
Proper coaching tool
Coaching style
(facing/non-facing)
Effectiveness Attitude change towards future
Promotion of self-directed
learning
Self-understanding and discovery
of strength
Voluntary goal setting and
execution
Reliability Ethical practice
Agreement and maintenance of
coaching relationships
Consistent sincerity
Forming and maintaining close
confidential relationship
Empathy Respect for client’s unique
individuality
Curiosity and focus on client
Positive attitude
Genuine and warm attitude
Responsiveness Flexible application of various
coaching methods
Positive feedback
Pacing such as breathing, speed
and tone
Exploring the other side and
reflecting meaning
32 J. Yang et al.

Table 2 Demographic characteristics of coaching expert

Items Sample size Ratio (%)
Age group Under 20 s 0 0
30 s 1 4.0
40 s 10 40
50 s 11 44
60 s 3 12
Coaching field Business coaching 3 12
Life coaching 17 68
Career coaching 1 4.0
Learning coaching 1 4.0
Etc 3 12
Coaching career Less than 3 years 4 16
3 years or more and less than 6 years 9 36
6 years or more and less than 9 years 5 20
9 years or more and less than 12 years 6 24
12 years or more 1 4.0
Total 25 100.0

3.3 Measurement Tool

This study used AHP technique to deduce determinants and relative importance
evaluation of coaching service quality, and designed the survey as paired comparison
as seen in Table 1. In order to secure accuracy of survey, the operational definition of
primary hierarchical factors of coaching service quality and example of measurement
were presented in Table 2. Survey questions were designed as paired comparison on
9-point scale, which is used in AHP technique, to calculate the importance.

4 Analysis and Results

4.1 Data Collection and Inspection

This study conducted survey from June 10 to June 20, 2020 (10 days) on 36 experts
who acquired professional coach certificate accredited by KCA (Korea Coach Asso-
ciation) and are working as a coach. Excluding 11 copies containing errors in the
analysis process, 25 respondents were selected.
Reliability of AHP technique is considered secured when the consistency ratio
(C.R) distinguishing error of respondent’s decision is 0.1 or below [32]. Among 36
collected samples, responses with C.R over 0.1 (11 copies) were eliminated, and
Analysis of the Relative Importance of Coaching Service Quality 33

Table 3 Consistency index by hierarchical factors

Primary Secondary hierarchical factors
hierarchical Expertise Tangibles Effectiveness Reliability Empathy Responsiveness
factors
C.R. 0.0172 0.0066 0.0283 0.001 0.0195 0.0063 0.0152

anal-ysis was performed on final 25 copies with secured reliability, in which the
verifica-tion result of C.R values of each hierarchical factor is displayed in Table 3.
Relative Importance Analysis Result Analysis result of relative importance of
coach-ing service quality is displayed in Table 4.
Based on analysis result of primary hierarchical factors, reliability (0.247) is the
most important factor, followed by expertise (0.202), empathy (0.200), responsive-
ness (0.168), effectiveness (0.140), and tangibles (0.043). In terms of the importance
of secondary hierarchical factors, the first factor of expertise is ability to listen active-
ly (0.443), followed by effective question (0.225), intuition and insight (0.216), and
coaching knowledge and experience (0.116). In terms of tangibles, proper place and
atmosphere for coaching (0.345) is the most important factor, followed by proper
coaching tool (0.274), coach’s attire and appearance (0.214), and coaching style
(0.167). Importance of effectiveness factors includes self-understanding and discov-
ery of strength (0.390), followed by voluntary goal setting and execution (0.256), atti-
tude change towards future (0.204), and promotion of self-directed learning (0.151).
Importance of reliability includes ethical practice (0.270), followed by form-ing
and maintaining close confidential relationship (0.258), consistent sincerity (0.249),
and agreement and maintenance of coaching relationship (0.223). Im-portance of
empathy factors includes genuine and warm attitude (0.283), followed by curiosity
and focus on client (0.263), respect for client’s unique individuality (0.243), and
positive attitude (0.212). Lastly, importance of responsiveness factors includes posi-
tive feedback (0.410), followed by exploring the other side and reflecting meaning
(0.250), pacing such as breathing, speed and tone (0.206), and flexible ap-plication
of various coaching methods (0.135).

5 Conclusions

Future society in the 4th Industrial Revolution Era will bring huge socioeconomic
changes based on AI, robotic technology, and biotechnology, and there will be con-
siderable changes in the form and properties of service [28]. Service economization
is a global trend, 70% of jobs are created in service industry, and the importance of
service industry is more significant than other industries in the 4th Industrial Revolu-
tion Era that the interest and demand in service are expected to grow further [29].
Coaching allows clients to have an insight and think through effective questions
in systematic conversations with processes. In other words, coaching is a very im-
portant industry in reinforcing competencies that humanity needs to possess in future
Table 4 The relative important of coaching service quality factors
34

Primary hierarchical factors Secondary hierarchical factors Combined importance

Factors Importance Ranking Factors Importance Ranking Importance Ranking
Expertise 0.202 2 Coaching knowledge and 0.116 4 0.0234 18
experience
Listening actively 0.443 1 0.0894 1
Effective question 0.225 2 0.0454 11
Intuition and insight 0.216 3 0.0436 12
Tangibles 0.043 6 Coach’s attire and appearance 0.214 3 0.0092 23
Proper place and atmosphere for 0.345 1 0.0148 21
coaching
Proper coaching tool 0.274 2 0.0117 22
Coaching style 0.167 4 0.0071 24
(facing/non-facing)
Effectiveness 0.140 5 Attitude change towards future 0.204 3 0.0667 17
Promotion of self-directed 0.151 4 0.0550 20
learning
Self-understanding and 0.390 1 0.0615 8
discovery of strength
Voluntary goal setting and 0.256 2 0.0637 15
execution
Reliability 0.247 1 Ethical practice 0.270 1 0.0285 3
Agreement and maintenance of 0.223 4 0.0211 7
coaching relationships
Consistent sincerity 0.249 3 0.0546 5
(continued)
J. Yang et al.
Table 4 (continued)
Primary hierarchical factors Secondary hierarchical factors Combined importance
Forming and maintaining close 0.258 2 0.0358 4
confidential relationship
Empathy 0.200 3 Respect for client’s unique 0.243 3 0.0486 10
individuality
Curiosity and focus on client 0.263 2 0.0526 9
Positive attitude 0.212 4 0.0424 13
Genuine and warm attitude 0.283 1 0.0566 6
Responsiveness 0.168 4 Flexible application of various 0.135 4 0.0228 19
coaching methods
Positive feedback 0.410 1 0.0693 2
Pacing such as breathing, speed 0.206 3 0.0348 16
and tone
Exploring the other side and 0.250 2 0.0422 14
reflecting meaning
Analysis of the Relative Importance of Coaching Service Quality
35
36 J. Yang et al.

society and adapt to future changes. Therefore, acknowledges the necessity to ex-
pand the base of coaching, therefore, develop the model of service quality determi-
nants in coaching service and analyze relative importance.
In the primary hierarchical factors, reliability (0.247) appeared as the most im-
portant factor. This is because the coaching relationship and effect of coaching var-
ies based on the formation of confidence with client. In counseling, formation of
relation with client is called rapport, and rapport is client’s feeling towards trust or
immersion to maintain relationship with the client. Rapport building induces positive
emotions in client and contributes to client satisfaction that it is used in mar-keting
as well. Coaching prioritizes in confidential relationship with client for an ef-fective
coaching session. Passmore [30] suggested 5 tasks for professional coaching activ-
ities, in which the third task is establishing a consensual code of conduct. Prior to
coaching, coach must participate in the consensual coaching relationship with client
and process with consistent sincerity and responsibility. Second factor in pri-mary
hierarchical factors of coaching service quality is expertise. Expertise is the skill or
competency that is the most importantly used in the interaction between coach and
client in coaching scene, which can be typically identified as active listening and
effective question. Previous studies on coaching competency include listening and
question in most competencies. Also, 11 core coaching competencies sug-gested by
ICF include active listening and powerful questioning as a part of effective communi-
cation skills. Through active listening, coaches listen to client’s nonverbal emotions,
suggestions, perception, value, interest, and belief, and allows client to be-come fully
aware of and express oneself by delivering the intended intention or key point. Based
on this, coaches asks effective questions so clients can pursue the de-sired target,
and client can have new and creative ideas, which can become the moti-vation of
growth and development. In this coaching process, coaches provide proper positive
feedback to client, and secondary hierarchical factor of responsiveness showed posi-
tive feedback (0.410) as the most important factor, as well as general importance
showing it as second important. Bandura [31] defines positive feedback as a catalyst
for successful performance of action for a specific goal. Therefore, positive feedback
such as recognition, support, and compliment in the conversation through listening
and question in coaching service helps client’s growth, and is very important in
coaching service quality.
Empathy factor (0.200) in primary hierarchy was the third important factor, and in
terms of sub-factors of empathy, sincere and warm attitude (0.283) appeared the most
important quality factor in coaching. DiDonato and Krueger [32] considers sincerity
as receptive ability and interprets it as dynamic status, not fixed, and Wick-ham [33]
states that people don’t become defensive and form trust when they recog-nize their
partner as sincere, and both parties can be connected stronger. Parasura-man et al.
[34] describes about inseparability in service quality, and defines that ser-vice does
not separate purchasing behavior and consumption behavior, but produc-tion and
consumption occur simultaneously. Therefore, coaching with service nature includes
sufficient connection and interaction with client through coach’s sincere attitude,
which in turn helps with client’s change and growth. Next, effectiveness (0.140) in
primary hierarchy is a consequential quality, which consists of criteria to recognize
Analysis of the Relative Importance of Coaching Service Quality 37

client’s change after coaching session, and includes 5 factors. Importance is relatively
lower than other quality factors, which can be interpreted as the confi-dence that
reliability, expertise, and empathy will positively lead to effectiveness of coaching. In
terms of importance of sub-factors of effectiveness, self-understanding and discovery
of strength (0.390) is the most important. Coaching stimulates client’s thinking and
brings new perspective and insight, so client discovers inner possibility and potential,
which accelerates client’s spontaneity and practice of self-directed life to set and
execute the goal. Therefore, coaching experts acknowledge self-understanding and
discovery of strength as the most important factor of the effec-tiveness of coaching.
If effects of postmodern era developed service industry, the upcoming 4th Indus-
trial Revolution Era is emphasizing the various forms and importance of service. In
addition, development and expansion of base of coaching industry is very important
to develop creative talent as humans are at the center of service, and quality of
coaching service needs to be improved as it can bring values to clients. Result of this
study was obtained from professional coaches, and is limited in interpreting client’s
expectation and opinion on coaching. Therefore, it should be expanded to clients
to measure and evaluate the coaching service quality. As one suggestion, coaching
knowledge and experience in sub-factors of expertise was the second-lowest in the
secondary hierarchy, and the 18th in overall importance. Coaching, with inseparable
characteristics, includes dynamic interaction that listening, questioning, and intuition
are more important that coaching knowledge and experience. However, there is no
academic system established for coaching yet. The first of 5 challenges suggested
by Passmore [30] is ‘explaining the system of coaching knowledge’. In order to
realize the identity and philosophy of coaching and establish it as a systemic field of
study, its knowledge system must be established first. Therefore, preparation for the
4th Industrial Revolution Era needs to be made by agreeing with the importance of
coaching knowledge and experience and establishing system as a field of study, and
expand the use of coaching to contribute to organizations and society.

References

1. Habermas J (1981) Modernity versus postmodernity. New German Critique 22:3–14

2. Choi I-S (2015) Critique on western-centrism inherent in post-modernism: focused on its
historical view. Korean Soc Polit Thought 21(2):99–121
3. McKnight S (2010) Spirituality in a postmodern age. Stone-Campbell J 13(2):211–224
4. Kim E-J (2018) The application effects of a coaching program on preservice childcare teachers.
J Korean Coach Res 11(3):73–91
5. Hamlin RG, Ellinger AD, Beattie RS (2009) Toward a process of coaching? A definitional
examination of “coaching”, “organization development”, and “human resource development.
Int J Evid Based Coach Mentor 7(1):13–38
6. Olorunniwo F, Hsu M, Udo G (2006) Service quality, customer satisfaction and behavioral
intentions in the service factory. J Serv Market 20(1):59–72
7. Asnawi A, Awang Z, Afthanorhan A, Mohamad M, Karim F (2019) The influence of hospital
image and service quality on patients’ satisfaction and loyalty. Manage Sci Lett 9:911–920
8. Berg ME, Karlson JT (2007) Mental models in project management coaching. Eng Manage J
19(3):3–13
38 J. Yang et al.

9. Nam SE, Ryu K (2017) An action research study of development and implementation of
coaching program for college students on academic probation (action research). Korean J
Gener Educ 11(1):281–311
10. Edwards L (2003) Coaching the latest buzzword or a truly effective management tool? Ind
Commerc Train 35(7):298–300
11. Whitmore J (2002) Coaching for performance. London: Nicholas Brealey Publishing
12. Stowell SJ (1988) Coaching: a commitment to leadership. Train Develop J 42(6):34–39
13. Miller L, Homan M (2002) Ace coaching alliances. Train Develop Am Soc Train Develop
56(1):40–46
14. ICF (2020) ICF. International Coach Federation. https://www.coachfederation.org
15. KCA (2020) KCA. Korea Coach Association. https://www.kcoach.or.kr/
16. Lee S, Kim K, Park H (2016) The teaching and training effect of the treasure talk coaching model
on the interaction of early childhood treasures and mothers. J Korean Coach Res 9(2):5–28
17. Fairley SG, Stout CE (2004) Getting started in personal and executive coaching: how to create
a thriving coaching practice. Wiley, Hoboken, NJ
18. Boyatzis AR (1982) The competent manager: a model for effective performance. Wiley, New
York, pp 20–21
19. Spencer LM, Spencer SM (1993) Competency at work. Wiley
20. Law H (2018) Groups and communities. Handbook of coaching psychology: a guide for
practitioners
21. Song YS (2011) A study on outside professional coach’s competencies for performance
improvement in corporate settings. Korean J Human Res Develop 13(1):53–73
22. Do MH, Kim SY (2019) Development and validation of a coaching competency scale. J Korean
Coach Res 12(4):59–81
23. Grönroos C (1984) A service quality model and its marketing implications. Eur J Mark
18(4):36–44
24. Parasuraman A, Zeithaml VA, Berry LL (1988) Servqual: a multiple-item scale for measuring
consumer perceptions of service quality. J Retail 64(Spring):12–40
25. Cronin JJ Jr, Taylor SA (1992) Measuring service quality: a reexamination and extention. J
Market 56:55–68
26. Shen XX, Tan KC, Xie M (2000) An integrated approach to innovative product development
using Kano’s model and QFD. Euro J Innov Manage 3(2):91–99
27. Etgar M, Fuchs G (2009) Why and how service quality perceptions impact consumer responses.
Manag Serv Qual Int J 19(4):474–485
28. Klaus S (2017) The Fourth Industrial Revolution. World Economic Forum
29. Byun DH (2018) Survey of service industry policy and big data analysis of core technology in
preparation of the fourth industrial revolution. J Serv Res Stud 8(1):73–87
30. Passmore J, Fillery-Travis A (2011) A critical review of executive coaching research: a decade
of progress and what’s to come. Coaching: An International Journal of Theory, Research and
Practice 4(2):70–88
31. Bandura A (1986) Social foundations of thought and action: a social cognitive theory.
Englewood Cliffs, NJ: Prentice-Hall
32. DiDonato TE, Krueger JI (2010) Interpersonal affirmation and self-authenticity: a test of
rogers’s self-growth hypothese. Self and Identity 9:322–336
33. Wickham RE (2013) Perceived authenticity in romantic partners. J Exp Soc Psychol 49:878–
887
34. PZB (1985) A conceptual model of service quality and Its implication for future research. J
Mark 49:41–50
Determinants of Technology-Based
Self-Service Acceptance

Seulki Lee and Donghyuk Jo

Abstract The spread of the COVID-19 has amplified the use of non-face-to-face
service due to the psychological risk of contact. As a result, non-face-to-face service
is being presented in a new paradigm that will lead the post-corona era, and the impor-
tance of this will be further increased. In particular, the service industry, which had
been face-to-face transaction, has made remarkable progress and is highly utilized
in the restaurant industry. Therefore, the purpose of this study is to explore the
factors that effects the acceptance intention of technology-based self-service of
food and beverage store customers. In the survey, we conducted a questionnaire
survey of customers who had used technology-based self-service at coffee shop.
Hypothesis test was done using AMOS statistical programs. The results of this study
have academic implications for investigating the antecedent variables to the accep-
tance intention of technology-based self-service in uncertain environments. In prac-
tice, it will contribute to decision-making and business strategy formulation for the
post-corona era.

Keywords TBSS (Technology-based self-service) · VAM (Value-based adoption

Model) · Perceived value · Attitude · Acceptance intention

1 Introduction

COVID-19 has influenced across the world on the way of life including the daily
routine of individual. For that, it serves as a motive that the face-to-face service
enterprise concentrates on the digitalization for the non-face-to-face service more
[1]. Particularly, the digitalization was accelerated in the food service area and retail
area, which is because of the fact that COVID-19 spreads by the human contact

S. Lee
Department of Business Administration, Sangmyung University, Seoul, Korea
e-mail: [email protected]
D. Jo (B)
Department of Business Administration, Soongsil University, Seoul, Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 39

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_4
40 S. Lee and D. Jo

and the social distancing campaign [2], for which the non-face-to-face consumption
culture is being spreading. Example of non-face-to-face consumption us self-service
encounter, online purchasing, online payment, kiosk, etc. [3].
The food service businesses and the retail businesses, which have invested proac-
tively in the non-face-to-face in our country, showed the pattern that the sales are
rather increased during the period that COVID-19 was spreading rapidly. Ono-face-
to-face order and payment system of Starbucks Korea can be the representative
example. Starbucks Korea is providing ‘Siren Order’, which is the non-face-to-face
service based on the technology that it has developed independently. ‘Siren Oder’ is
the service that the customer can receive the beverage personally after selecting and
paying the desired beverage through dedicated APP and can minimize the contact
with the employee. The number of order received through ‘Siren Order’ from January
to February, 2020 was more than 8 million and increased by 25% compared to the
same period of 2019 [4].
‘Siren Order’ of Starbucks Korea is a sort of TBSS (Technology-based Self-
service) and corresponded to Smart Ordering System. TBSS means the technology
service that the consumer can participate directly to the service production through
machine or internet, etc. without help of employee [5]. Recently, in the food and
beverage store, Smart Ordering System is utilized actively [6]. Smart Ordering
System has advantage that it takes short time from ordering to payment through
the mobile devices possessed by the individual and can minimize the communica-
tion error. However, in the position of business, since the initial cost is incurred by
the customized software development by enterprise and authentication [6] and there
is no installation-related cost transferred to consumer, if the user wants, the user can
stop using the service any time.
Therefore, to TBSS like smart ordering system, forecasting the intention if the
user would use relevant technology is essential. In addition, various researches on the
kiosk that minimizes the contact through the device installed in specific location have
been made but the research on the smart ordering system, which TBSS is realized
through mobile devices carried by user is not enough.
Therefore, the purpose of this study is to verify the factors having impact on the
TBSS acceptance intention empirically focusing on the smart ordering system. To
do that, this study established the research model based on the value-based adoption
model saying that considering the technology user as consumer, the value perceived
from technology would increase the technology acceptance intention, and to examine
the relations by drawing key factors based o preceding research. In the meantime,
since most of the smart ordering systems in the food and beverage store do not incur
the initial installation cost in the user, have been introduced for the convenience of
use and its technology is not complicate, the key factor was limited to the benefits.
The results of this study are expected to provide the information what are the
factors that the users consider as important to the organization, which wants to
introduce the smart ordering system as a part of TBSS at the time of preparing
post COVID era.
Determinants of Technology-Based Self-Service Acceptance 41

2 Theoretical Background and Hypotheses

2.1 TBSS (Technology-Based Self-Service)

TBSS means the service that allows the consumer to participate directly in the service
production through the machine or technology such internet, etc. without help of
employee [5]. Touch screen, self-scanning, ATM, internet banking, mobile banking,
chatbot, etc. are considered as TBSS. Recently, as the mobile devices like smart
phone become importance interface of TBSS by the increase of availability of Wi-
Fi, Bluetooth, etc., new mobile-based TBSS is emerged [7]. Therefore, in the food
service industry, the mobile-based smart ordering system is included as a type of
TBSS [6].
Smart ordering system composed of 3 elements. First, the user must have Web
Ordering System in his/her smart phone. Next, it should have Menu Management
System to be able to select menu and Order Retrieval System that the employee can
see the order details ordered by user [8]. As such, the smart ordering system has a
hassle that the user must install the system in the personal mobile device.
Nevertheless, the reason why the user prefers to TBSS like smart ordering system
is because of the speed and convenience taken to make order, cost saving, fun,
minimization of communication error, minimization of unnecessary meeting with
employee, etc. [5, 6]. In our country, “Siren Order’ of Starbucks can be the represen-
tative example of smart ordering system. ‘Siren Order’, which was TBSS developed
independently by domestic Starbucks, is the service that the customer can select and
receive the desired beverage after payment through dedicated APP. Particularly, since
it is designed to select the personal option for product, the individual customized
service is possible. In addition, through the APP notice, the information such as
various event benefits, promotional beverage, etc. are provided [9]. Therefore, in this
study, the factors having influence on the mobile device-oriented TBSS acceptance
intention will be examined using Siren Order of Starbucks.

2.2 VAM (Value-Based Adoption Model)

As the representative theory on the process that the user is accepting specific tech-
nology, there is TAM (Technology Acceptance Model) suggested by Davis et al.
[10]. However, TAM has limitation that it could not include the general consumer
as a subject of accepting technology. To overcome such limitation, VAM (Value-
based Adoption Model) was emerged trying to explain by what the service consumer
accepts the technology [11]. VAM defined the IT technology user as consumer and
focused on maximizing the personal value of the consumer [11]. Kim et al. [11] tried
to explain the acceptance of technology and service based on the concept of perceived
value by Zeithaml [12]. Therefore, the perceived value from the new technology
becomes importance antecedent for technology acceptance intention.
42 S. Lee and D. Jo

In the meantime, the consumer experiences the benefit and sacrifice obtained by
using IT technology before perceiving comprehensive value. Therefore, both the
benefit and sacrifice should be considered as an element having influence on the
value in accepting the technology and in VAM, the usefulness and enjoyment were
suggested as element of benefit and the technicality and perceived fee were as the
element of sacrifice [11, 13].
The usefulness means the degree that new technology is useful in the daily life
or business, etc. and it is the similar concept to the functional advantage of product
or service use [11, 14]. The usefulness of the technology can improve the ability of
carrying out the work and makes the decision making easy [14]. Therefore, the more
the new technology is perceived as useful, the higher the value of the technology
can be perceived, which is proved in the research related to the multiple technology
acceptance [11, 14, 15].
Enjoyment means the degree of pleasure, joy and satisfaction felt using new
technology [10, 11]. The person who experiences the pleasure and joy by using the
technology is highly likely to accept the technology and use widely by perceiving
the value of the technology highly [10]. Therefore, the multiple researches utilize
the enjoyment as major predictor of the perceived value [15–17].
Technicality is the degree of perceiving technically superior in the process of
providing services and is determined by the system reliability, connectivity and effi-
ciency [11]. If in the position of technology user, the use of technology is felt difficult
and the time to connect it is taken longer, it is recognized as cost and causes the nega-
tive mental state. However, on the contrary, if the use of technology is not complicate
and easy, the positive value can be formed. The research results that the complexity
of the technology has negative influence on the perceived value supports it [15].
Perceived cost is paying the monetary consideration, which includes the actual
price of the product or service, and measured with the consumer’s perception on
the cost actually paid [11]. If the cost paid to use the technology is perceived high,
negative perception can be formed. On the contrary, the more the cost is saved or
additional benefit is increased, the perceived value can be increased [18]. As the
smart ordering system, the object of this study, does not generate the cost related to
use of system besides the existing price of product ad provides the benefit such as
accumulation of point, issuance of coupon, etc., it will be used by changing to cost
advantage. In the meantime, the research by Kwon and Seo [19] showed that the
economic feasibility has positive influence on the perceived value.
Through the aforementioned contents, it was expected that the usefulness, enjoy-
ment, technicality and the cost advantage would have influence on the perceived value
of the technology-based self-service in the food and beverage store and following
hypothesis was drawn.
H1. Perceived benefit will have a positive effect on the perceived value.
Perceived value means overall evaluation on the product or service utility [12] and
is emphasized as a key antecedent determining the consumer’s attitude or behavior
[20]. For example, the research that explains the acceptance of the service combined
with new technology such as mobile hotel reservation system [21], IPTV [22], e-
book subscription service [23], IoT smart home service [24], VR [15], etc. proved
Determinants of Technology-Based Self-Service Acceptance 43

that the perceived value is important predictor of acceptance intention. Therefore, it

can be expected that in case of the user perceiving the value of TBSS, the acceptance
intention would be increased and following hypothesis was established.
H4. Perceived value will have a positive effect on acceptance intention.

2.3 VAB (Value-Attitude-Behavior) Model

As the model predicting consumption behavior of the consumer, there is VAB (Value-
Attitude-Behavior) Model. VAB Model suggested by Homer and Kahl [25] is the
model useful in predicting the consumption behavior and is the theory that the
consumer’s behavior is formed through the value and attitude.
Attitude in VAB Model plays the mediating role between the abstract value and
specific behavior [25]. Attitude is the learned emotional inclination relatively formed
consistently on the specific target and means the psychological reaction appeared
favorably or unfavorably through the experience [26]. Dabholka and Bagozzi [27]
defined the attitude as the experiential evaluation such as good or bad, pleasant or
unpleasant, etc. Therefore, in order to form the attitude, the cognitive reaction like
value perception should be preceded.
Since the value means overall evaluation on the utility of product or service
according to Zeithaml [12], it is viewed as single dimension but is viewed as multi-
dimensional perspective depending on the researcher [28]. While the approach to
the single dimension has overall character of practical aspect, in case of approaching
to multi-dimension, it has strong complex characters of psychological, cognitive
and emotional dimension [29]. In the multi-dimensional structure, utilitarian value,
hedonic value, acquisition value, transaction value, use value, functional value,
emotional value, social value, etc. are used [28, 30, 31]. If multi-dimensional struc-
ture is examined, it is observed that the aspect of diverse benefits obtained by the
experience is considered.
Therefore, it can be expected that the benefit element of technology use drawn
according to VAM (usefulness, enjoyment, technicality, cost advantage) also have
significant influence on the attitude. Gan and Wang [32] examined the influence
on the attitude by establishing the benefits of social commerce as value. Therefore,
following hypotheses were drawn.
H2. Perceived benefit will have a positive effect on the attitude.
H3. Perceived value will have a positive effect on the attitude.
The attitude influenced by the value influences on the consumer’s behavior [24].
Dabholka and Bagozzi [27] verified that the behavior life decision making is influ-
enced by the attitude. Karjaluoto et al. [33] suggested in the research on the mobile
banking acceptance intention that the attitude can become a powerful antecedent, and
Alotaibi [34] proved that the attitude has influence on the cloud computing accep-
tance. The research on Home Smart IoT service by Kim et al. [24] demonstrated
that the attitude is the major influential factor of the acceptance intention, through
44 S. Lee and D. Jo

Table 1 Sample characteristics

Category and items Sample size Ratio (%)
Gender Female 127 53.8
Male 109 46.2
Age 20–29 74 31.4
30–39 109 46.2
40–49 40 16.9
More than 50 13 5.5
Frequency of use (recent 6 month) 1–2 100 42.4
3–4 46 19.5
5–6 26 11.0
More than 7 64 27.1

which it was inferred that the attitude would improve the intention of accepting new
technology and following hypothesis was drawn.
H5. Attitude will have a positive effect on the acceptance intention.

3 Research Method

3.1 Sample and Data Collection

In order to verify the proposed model, we conducted a questionnaire survey for

customers who have used the smart order system (siren order) in Starbucks Korea.
As a result of the survey, 250 questionnaires were collected, and 236 cases were
selected as valid samples after eliminating missing or inadequate data. The samples
of this study are summarized in Table 1.

3.2 Measures

For the content validity of the measurement tool, measurement items that have already
been verified in previous studies were derived and modified to suit the purpose of
this study. In this study, usefulness, enjoyment, technology, cost advantage, perceived
value, attitude, and acceptance intention were set as the main concepts. For tools to
measure the concept, studies by Kim et al. [11], Lin et al. [22], Yu et al. [14], and Kim
et al. [24] were referenced, and then measured them by the Likert Five-point scale
(from Not at all to Very Much). The measurement items in this study are summarized
in Table 2.
Determinants of Technology-Based Self-Service Acceptance 45

Table 2 Confirmatory factor analysis based on reliability

Variable Measurement items Factor L.D. C.R. Crb. Alpha
Usefulness Reduction of time required 0.783 0.944 0.893
Helped to achieve the desired 0.845
purpose
Usefulness of the order 0.836
method
Usefulness of payment 0.836
method
Enjoyment Amusement 0.882 0.897 0.895
Attractive 0.824
Pleasure 0.876
Technicality Ease of use 0.798 0.914 0.868
Available anywhere 0.861
Less occurrence of problems 0.829
Cost advantage Economic benefits 0.793 0.898 0.811
Reasonable price 0.864
Perceived value Value for money invested 0.828 0.921 0.879
Benefits for the effort invested 0.837
Overall value 0.856
Attitude Prefer ordering through Siren 0.643 0.901 0.770
order
Satisfaction with the services 0.787
provided
Overall positive experience 0.767
Acceptance intention Want to use it in the future 0.836 0.903 0.826
Will be used when ordering in 0.793
the future
Want to use even if there are 0.722
competitors

3.3 Analysis Method

For hypothesis testing, we first carried out measurement model analysis for validity
and reliability analysis, and then structural equation model analysis using Amos
v.22.0 program.
46 S. Lee and D. Jo

Table 3 Discriminant validity

Variable 1 2 3 4 5 6 7
1. Usefulness 0.899a
2. Enjoyment 0.115 0.862a
3. Technicality 0.093 0.196 0.884a
4. Cost advantage 0.491 0.134 0.166 0.903a
5. Perceived value 0.393 0.217 0.564 0.489 0.892a
6. Attitude 0.490 0.150 0.411 0.508 0.596 0.868a
7. Acceptance intention 0.406 0.176 0.486 0.493 0.649 0.579 0.871a
Note a square root of AVE (Average Variance Extract)

4 Analysis and Results

4.1 Measurement Model

Confirmative factor analysis was conducted to secure the reliability and validity of the
measurement tool. For this, standard χ2 (χ2 /df ), RMSEA, GFI, NFI, TLI, CFI were
used to check goodness of fit. As a result of confirmatory factor analysis of measure-
ment model, χ2 /df = 1.21(≤3), RMSEA = 0.03(≤0.05), GFI = 0.927(≥0.9), NFI
= 0.934(≥0.9), TLI = 0.985(≥0.9), CFI = 0.988(≥0.9) all indicators were found
to be suitable. After verifying measurement model’s fitness, reliability and validity
were analyzed. Reliability is examined by accessing construct reliability (C.R.). C.R.
ranges from 0.897 to 0.944, exceeding the cut-off value of 0.7 for C.R. Validity is
was evaluated by average variance extracted (AVE), and it was found to be 0.744–
0.815, exceeding the standard value of 0.5. Additionally, to identify the discriminant
validity, the square root of the AVE for each construct is compared with the correlation
coefficients between two constructs [35, 36]. When the AVE square root is larger
than the correlation coefficient, it is determined that the discrimination validity is
secured. As a result of analysis, reliability and validity were verified and the detailed
results are presented in Tables 2 and 3.

4.2 Structural Model

The fitness of the measurement model and the reliability and validity of the measure-
ment tool were confirmed to be normal, and the structural model was analyzed. As a
result of structural model’s fitness test, χ2 /df = 1.225(≤3), RMSEA = 0.031(≤0.05),
GFI = 0.925(≥0.9), NFI = 0.932(≥0.9), TLI = 0.984(≥0.9), CFI = 0.987(≥0.9),
all of which were above the baseline.
Determinants of Technology-Based Self-Service Acceptance 47

4.3 Hypotheses Tests

Structural equation model analysis was performed to verify the hypothesis of the
proposed model. As a result, first, Usefulness (β = 0.149, C.R. = 2.210), Technicality
(β = 0.548, C.R. = 8.572) and Cost advantage (β = 0.386, C.R. = 5.518) have a
positive effect on Perceived value, therefore, H1a, H1c and H1d were supported.
On the other hand, Enjoyment (β = 0.047, C.R. = 0.887) did not have a positive
effect on Perceived value, therefore, H1b was not supported. Second, Usefulness (β
= 0.265, C.R. = 3.425), Technicality (β = 0.246, C.R. = 2.631) and Cost advantage
(β = 0.278, C.R. = 2.903) have a positive effect on Attitude, therefore, H2a, H2c
and H2d were supported. However, Enjoyment (β = -0.018, C.R. = -0.315) did
not significantly affect Attitude, therefore, H2b was not supported. Third, perceived
value(β = 0.289, C.R. = 2.422) was shown to have a positive effect on attitude, and
Hypothesis 3 was supported. Fourth, perceived value(β = 0.482, C.R. = 4.875) was
shown to have a positive effect on acceptance intention, therefore H4 was supported.
Finally, it was confirmed that attitude(β = 0.401, C.R. = 3.864) had a significant
positive(+) effect on acceptance intention. Thus, H5 was supported. The results of
hypotheses test are summarized in Table 4.

Table 4 Results of hypotheses tests

H Path Estimate (β) C.R. (t) Result
H1a Usefulness → Perceived value 0.149 2.210* Supported
H1b Enjoyment → Perceived value 0.047 0.887 Not supported
H1c Technicality → Perceived value 0.548 8.572*** Supported
H1d Cost advantage → Perceived value 0.386 5.518*** Supported
H2a Usefulness → Attitude 0.265 3.425*** Supported
H2b Enjoyment → Attitude −0.018 −0.315 Not supported
H2c Technicality → Attitude 0.246 2.631** Supported
H2d Cost advantage → Attitude 0.278 2.903** Supported
H3 Perceived value → Attitude 0.289 2.422* Supported
H4 Perceived value → Acceptance intention 0.482 4.875*** Supported
H5 Attitude → Acceptance intention 0.401 3.864*** Supported
*p < 0.05; **p < 0.01; ***p < 0.001
48 S. Lee and D. Jo

5 Conclusions

5.1 Summary and Discussion of Results

This study intended to draw the influential factor based on the VAM (Value-based
Adoption Model) and VAB (Value-Attitude-Behavior) model and to identify the
structural relations among variable in order to identify the factors of promoting the
user’s acceptance intention of TBSS, which is receiving the attention by the risk
recognition on the contact due to COVID 19 virus. Particularly, Smart Ordering
System, which is being introduced actively in the store specialized in beverage of
food service sector was examined intensively. Therefore, the study was performed
focusing on ‘Siren Order’ of Starbucks Korea, which was developed independently
by the enterprise and already has multiple users. The summary of research results
are as follows.
However, the enjoyment appeared not to have significant influence on the
perceived value and the attitude, which means the more it is believed that Smart
Ordering System would help the user in the life by using it, the easier how to use and
the less failure and the more it is perceived that the economic benefits are provided,
the more the value perceived comprehensibly and the attitude toward the technology
are improved. Second, it was confirmed that the perceived value has significant posi-
tive (+) effect on the attitude, which means that the better the value of new technology,
the more positive attitude is formed. Third, as the perceived value and the attitude
appeared to have significant positive (+) effect on the acceptance intention, it supports
the results of preceding researches [15, 21–24, 33, 34].

5.2 Implications and Limitations

The academic implications of this study are as follows. First, in the circumstance that
the empirical research on Smart Ordering System, a kind of TBSS, is not enough, the
theoretical system, that explains the intention that the user would accept the relevant
technology, was prepared. Second, this study has meaning that the influential rela-
tions among the concepts were identified by applying VAM (Value-based Adoption
Model), which examines the influential factor in the aspect of perceived value by the
user in accepting new technology, and VAB (Value-Attitude-Behavior) Model, which
the consumer’s behavior is receiving the influence by the value and the attitude.
Practical implications are as follows. First, from the fact that the usefulness,
technicality, cost advantage suggested as benefit by the technology in VAM have
positive influence on the perceived value and the attitude toward Smart Ordering
System, the need of design to be able to emphasize the usefulness, technicality,
cost advantage is suggested. For example, diverse methods are introduced in order to
satisfy the individual need in the ordering method and payment method through Smart
Ordering System and at the same time, the operating method should be designed
Determinants of Technology-Based Self-Service Acceptance 49

conveniently. In addition, considering that the cost advantage has significant influence
on the value perception and attitude formation, it is necessary to perform the service
or promotion that can provide the monetary benefits only from relevant system.
In case of “Siren Order’ of Starbucks Korea, it provides the benefit in the aspect
of cost to user through the issuance of BOGO coupon (1 + 1), the free beverage
through the accumulation of Star, performance of specific event, etc. Second, since
the enjoyment, which was proved in multiple researches, represented not to have
significant influence on the perceived value and the attitude, it can be estimated that
it would not be the major factor in the smart ordering system, of which purpose of
use is simple and clear. Therefore, if the food service sector would introduce smart
ordering system, it is necessary to establish the strategy considering the aspect of
usefulness, technicality and cost advantage with priority than the elements related
to the enjoyment. Third, it was confirmed that the more the perceived value and the
attitude toward the technology are positive, the higher the user’s intention to accept
the smart ordering system out of TBSS, through which the need to seek out the
measures to induce the acceptance can be suggest by stimulating the aspect of user’s
value and attitude.
The limitations of this study and the direction of follow-up research are as follows.
First, to understand the non-face-to-face consumption culture at the point preparing
post-Covid-19 era, the aspects of benefit and value received from new technology
were examined from the smart ordering system, which is a kind of TBSS. However,
the innovation of individual makes to take more active action in accepting new
technology [14]. Therefore, it is regret that in this study, the innovative disposition
possessed by the individual user was not considered, and in the follow-up research,
the characteristics of the individual before contacting with new technology should
be considered. Second, since this study was performed focusing on ‘Siren Order’ of
Starbucks Korea, which is the Smart Ordering System introduced for the first time in
the domestic food and beverage store, it is necessary to make effort to generalize the
research results by performing the comparison with the system introduced by other
brand or the integrated research.

References

1. Diebner R, Silliman E, Ungerman K, Vancauwenberghe M (2020) Adapting customer

experience in the time of coronavirus. McKinsey & Company. Accessed 19 May 2020
2. Shahbaz M, Bilal M, Moiz A, Zubair S, Iqbal HM (2020) Food safety and COVID-19: precau-
tionary measures to limit the spread of coronavirus at food service and retail sector. J Pure Appl
Microbiol 14(Special Edition):1–9
3. Lee SM, Lee D (2020) “Untact”: a new customer service strategy in the digital age. Serv Bus
14(1):1–22
4. Starbucks Korea. http://starbucks.co.kr/bbs/getBodoView.do?seq=3767 (2020)
5. Meuter ML, Ostrom AL, Roundtree RI, Bitner MJ (2020) Self-service technologies: under-
standing customer satisfaction with technology-based service encounters. J Market 64(3):50–64
6. Umap S, Surode S, Kshirsagar P, Binekar M, Nagpal N (2018) Smart menu ordering system
in restaurant. IJSRST 4(7):207–212
50 S. Lee and D. Jo

7. Tanimura M, Ueno T (2013) Smartphone user interface. Fujitsu Sci Tech J 49(2):227–230
8. Mishra N, Goyal D, Sharma AD (2018) Automation in restaurants: ordering to robots in
restaurant via smart ordering system. Int J Technol Manage 4(1):1–4
9. Baek SH, Yim HR, Kim HS (2019) A study of the acceptance intention of customers in coffee
franchise using extended technology acceptance model (E-TAM): focused on starbucks mobile
application. Culinary Sci Hospitality Res 25(4):146–155
10. Davis FD, Bagozzi RP, Warshaw PR (1989) User acceptance of computer technology: a
comparison of two theoretical models. Manage Sci 35(8):982–1003
11. Kim HW, Chan HC, Gupta S (2007) Value-based adoption of mobile internet: an empirical
investigation. Decis Support Syst 43(1):111–126
12. Zeithaml VA (1988) Consumer perceptions of price, quality, and value: a means-end model
and synthesis of evidence. J Market 52(3):2–22
13. Chunxiang L (2014) Study on mobile commerce customer based on value adoption. JApSc
14(9):901–909
14. Yu J, Lee H, Ha I, Zo H (2017) User acceptance of media tablets: an empirical examination of
perceived value. Telematics Inform 34(4):206–223
15. Vishwakarma P, Mukherjee S, Datta B (2020) Travelers’ intention to adopt virtual reality: a
consumer value perspective. J Destination Mark Manage 17:1–13
16. Yang H, Yu J, Zo H, Choi M (2016) User acceptance of wearable devices: an extended
perspective of perceived value. Telematics Inform 33(2):256–269
17. Kim SH, Bae JH, Jeon HM (2019) Continuous intention on accommodation apps: integrated
value-based adoption and expectation-confirmation model analysis. Sustainability 11(6):1578
18. Lovelock CH (2001) Services marketing. Prentice Hall, Upper Saddle River, NJ
19. Kwon HK, Seo KK (2013) Application of value-based adoption model to analyze SaaS adoption
behavior in Korean B2B cloud market. Int J Advancements Comput Technol 5(12):368–373
20. Tam JL (2004) Customer satisfaction, service quality and perceived value: an integrative model.
J Mark Manage 20(7–8):897–917
21. Wang HY, Wang SH (2010) Predicting mobile hotel reservation adoption: insight from a
perceived value standpoint. Int J Hospitality Manage 29(4):598–608
22. Lin TC, Wu S, Hsu JSC, Chou YC (2012) The integration of value-based adoption and expec-
tation–confirmation models: an example of IPTV continuance intention. Decis Support Syst
54(1):63–75
23. Hsiao KL, Chen CC (2017) Value-based adoption of e-book subscription services: the roles of
environmental concerns and reading habits. Telematics Inform 34(5):434–448
24. Kim Y, Park Y, Choi J (2017) A study on the adoption of IoT smart home service: using
value-based adoption model. Total Qual Manage Bus Excellence 28(9–10):1149–1165
25. Homer PM, Kahle LR (1988) A structural equation test of the value-attitude-behavior hierarchy.
J Pers Soc Psychol 54(4):638–646
26. Aaker JL (1997) Dimensions of brand personality. J Mark Res 34(3):347–356
27. Dabholkar PA, Bagozzi RP (2002) An attitudinal model of technology-based self-service:
moderating effects of consumer traits and situational factors. J Acad Mark Sci 30(3):184–201
28. Sweeney JC, Soutar GN (2001) Consumer perceived value: the development of a multiple item
scale. J Retail 77(2):203–220
29. Sanchez J, Callarisa L, Rodriguez RM, Moliner MA (2006) Perceived value of the purchase
of a tourism product. Tour Manag 27(3):394–409
30. Peña AIP, Jamilena DMF, Molina MÁR (2012) The perceived value of the rural tourism stay
and its effect on rural tourist behaviour. J Sustain Tour 20(8):1045–1065
31. Shan G, Yee CL, Ji G (2020) Effects of attitude, subjective norm, perceived behavioral control,
customer value and accessibility on intention to visit Haizhou Gulf in China. J Mark Adv Pract
2(1):26–37
32. Gan C, Wang W (2017) The influence of perceived value on purchase intention in social
commerce context. Internet Res 27(4):772–785
33. Karjaluoto H, Püschel J, Mazzon JA, Hernandez JMC (2010) Mobile banking: proposition of
an integrated adoption intention framework. Int J Bank Market 28(5):389–409
Determinants of Technology-Based Self-Service Acceptance 51

34. Alotaibi MB (2014) Exploring users’ attitudes and intentions toward the adoption of cloud
computing in Saudi Arabia: an empirical investigation. J Comput Sci 10(11):2315–2329
35. Fornell C, Larcker DF (1981) Evaluating structural equation models with unobservable
variables and measurement error. J Mark Res 18(1):39–50
36. Ping RA Jr (2004) On assuring valid measures for theoretical models using survey data. J Bus
Res 57(2):125–141
Pre-verification of Data in Electronic
Trade Blockchain Platform

Saeyong Oh, Sanghyun Cho, Sunghwa Han, and Gwangyong Gim

Abstract Blockchain technology has several advantages, including security, trans-

parency, and stability. It can reduce data management cost and be applied to impor-
tant electronic document related services. However, in the perspective of the subject,
in order to apply the blockchain technology to electronic document-related services
form, it is necessary to review the integrity of the electronic documents to verify that a
stable transaction is possible. In addition, in the perspective of the consumer, because
electronic documents contain content related to their important personal information
or contracts, electronic documents should be proved to be stable in the blockchain
trading platform. Therefore, this study aims to present an architecture that can verify
the integrity of the blockchain-based electronic document platform by presenting
a pre-validation method for electronic documents. Furthermore, by establishing an
actual pilot system based on this architecture, a system where customers can reliably
understand their transactions and contracts was made.

Keywords Blockchain · Electronic trade platform · Data pre-verification · Data

efficiency · Data integrity · File format verification · Data forgery · Electronic
document

S. Oh · S. Cho · S. Han · G. Gim (B)

Department of IT Policy and Management, Soongsil University, Seoul, South Korea
e-mail: [email protected]
S. Oh
e-mail: [email protected]
S. Cho
e-mail: [email protected]
S. Han
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 53

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_5
54 S. Oh et al.

1 Introduction

The scalability of the blockchain technology can be seen as a technology that can
solve the problem of efficiency and reliability of the storage and management of elec-
tronic document data such as various records, medical data, and contract documents
in daily life [1–3].
Electronic documents are typically stored and managed by the entity providing the
services, and this centralized structure is a significant burden to information manage-
ment entities such as businesses and public institutions. This is because the volume
of electronic documents increases in the contemporary society and the cost of solving
security problems increases. Distributed network systems in blockchain technology
can be the next-generation core technology to address these costs [3–5]. There are
many things to consider when the provider of e-document-related services decide to
provide it services based on the blockchain technology. The primary provider should
validate the existing electronic document data. Blockchain cannot be deleted or
modified after upload, so the information subject must be able to determine whether
electronic documents are valid data before the data is transferred. Next, the integrity
of electronic documents on the blockchain should be checked to verify that forged
data are negotiable in the distribution of electronic documents. Finally, a system that
can check transaction details or transaction information should be established so that
users can have faith in the blockchain technology.
This study attempts to experiment with the above three factors and suggest solu-
tions. First, for verification of validity of electronic documents, we would like to
select the data format that accounts for the majority of electronic documents. Also,
we will suggest a method to verify format verification through the format analyzer
of electronic documents, and to verify the damage of files using the data validation
model. Second, blockchain is said to be theoretically impossible to falsify data and
safe from hacking threats. However, there has been no research that has actually
verified it with electronic document files. Therefore, it is necessary to verify various
variables of electronic document forgery and verify them in the blockchain. To do
so, we would like to check the integrity of the data by selecting a blockchain that can
be applied to electronic transactions and verifying it experimentally. Third, because
the blockchain is in code form, it is difficult for ordinary users to use it, so we would
like to create a pilot system that looks like a service portal or transaction system by
visualizing transaction details or contracts [4, 6, 7].

2 Theoretical Background

2.1 Advent and Features of Blockchain

A blockchain can be defined as a block configuration that holds records of data in

the form of a bundle of individual transactions in a distributed transaction. The block
Pre-verification of Data in Electronic Trade Blockchain Platform 55

maintains a growing list of data records protected from tampering and forgery. Each
of these blocks includes a timestamp and a link to the previous block [8, 9, 10].
In short, the blockchain is a network-based distributed system [11, 12]. The most
distinctive feature of blockchain is that it is a distributed book-operating structure
without a central management entity acting as a broker for the transactions. It has
a structure that directly transmits and receives data and shares related information
between each node participating in the network without a central control system.
This structure eliminates centralization of data, maintains security through consensus
algorithms such as strong encryption and proof of work, and distributes records across
all nodes in the network, enabling transparency in transactions and stable operation
of the system [4, 13, 14].

2.2 Structure of Blockchain

Blockchain can literally be described with blocks and chains. Each block records the
data exchanged between all network participants and each other, and these blocks
are linked by a chain. Most transaction-related information is contained in the Trans-
action Details section. Any change in transaction information will result in different
hash values [15]. Changes can be easily detected if a block of data is arbitrarily
modulated. The block contains the hash number of the previous block as well as
its own hash value. Because of this connection, blocks are linked together to form
chains, and the longer the chain is, the more difficult it is to forge [16–18].

2.3 Advanced Research

Blockchain technology has the scalability to be utilized in a variety of areas [4,

19, 20]. This section provides an overview of the active research using blockchain
technology in diverse fields. First, Civelek and Ozalp [21] stated that research on
blockchain technology and paperless trade proves that the blockchain technology
can provide document integrity in foreign trade and finance. The study argued that
changes such as existing legal problems and infrastructure should be integrated into
the platform.
Ghazali and Saleh [22] designed a model that verifies the certificate by requesting
it to a public infrastructure-based blockchain system by generating hash values using
hash algorithms and generating private keys and timestamps in order to give it an
essential element of the blockchain. Through this, the author suggested the possibility
that blockchain technology can be used for degree certification.
Nizamuddin et al. [23] proposed a smart contract code solution to avoid security
hacking by developing a solution to store documents in distributed systems and
control versions of documents in distributed document version control studies using
Ethereum blockchain and IPFS (InterPlanetary File System).
56 S. Oh et al.

3 Empirical Analysis

3.1 Efficiency of Pre-verification Model for Electronic

Document Data

In the stage of verifying the validity, two stages of experiment were done to verify the
file format of electronic document and the validity of the file itself. The experiment
used language C/C++, and in the text verification stage, because there are many text
format types, it is hard to verify all the documentation formats. But the experiment
tried to embody test model that can verify a broader range of formats. Even if a
document format is distinguished and verified as a document that can be opened
properly at the stage of verifying the validity of documents, because there is no
method to open the damaged bit stream or encrypted electronic documents, this part
was exempted in the experiment.
Document File Format Verification
The e-document format analyzer built in this study was targeted at the following
Hancom format, MS office-related format, and image format, and was conducted
by extracting identifiable information from the headers of the text and attachments
in the document to verify the format and comparing it with the DB containing the
format information.
As seen in Table 1, each file format was normally recognized.
Document’s Efficiency Verification
The file’s efficiency verification tool developed in relevant study is designed to
allow the file format to perform an effectivity verification only on the certified format,
and the efficiency verification is possible if the added file format can be verified in
the file format analyzer. The tool verifying file’s efficiency is configured to check if
the file is damaged, and the file’s damage types are mainly as follows.
• File of which the name exists but content is empty.
• Documents that cannot be opened due to partial damage.
• Documents that cannot be opened because there is a password on the file.
• Document that cannot open file because file header is damaged.
As seen above there were some files that were damaged. These files were judged
to not be worth keeping it because there is no way to open it, and this is the reason

Table 1 The result of

Division File format(extension) Test result
document format verification
experiment Hangeul hwp(3.0, 4.0, 5.0) Normal
MS OFFICE xlsx, xls, docx, doc, pptx ppt Normal
Image png, jpeg, gif Normal
PDF PDF Normal
Attached file hwp, doc, pdf Normal
Pre-verification of Data in Electronic Trade Blockchain Platform 57

Fig. 1 Efficiency
verification algorithm of
electronic document data

why file’s efficiency verification is needed. For the design of the file’s efficiency
verification tool, the verification tool was designed based on the architecture as
shown in Fig. 1, and its configuration is as follows.
The experiment was conducted on the electronic document validation tool, and
the sampled text document selected 50 files from the document format analyzer, of
which 20 files were selected, each with an empty file, a file with a damaged document,
a file with a password, and a file with a damaged document header. The remaining
files were separated into normal documents to validate the electronic document.
As Table 2 shows, the test confirmed that all 50 files were normally validated.
The empty files and password-set files, and the document header showed accurate
classification results for five corrupted files and parts of the files, and all the other
normal files were classified as normal.

3.2 The Result of Integrity Verification on Electronic

Document Data

To verify the integrity of the blockchain of electronic documents, the appropriate

blockchain source for electronic document transactions was selected and the test
environment was composed of test servers and nodes based on the selected source.
58 S. Oh et al.

Table 2 Document’s efficiency verification

Division File name Extension File size Whether test Verification Verification
(KB) data error exists result contents
1 Hangeul1 hwp 15 Normal Success Normalfile
2 Hangeul2 hwp 20 Normal Success Normalfile
3 Hangeul3 hwp 15 Error Empty Empty
document document
detection
4 PDF1 pdf 30 Normal Success Normalfile
5 PDF2 pdf 30 PASSWORDS passwords Passwords
document
detection
6 PDF3 pdf 50 Normal Success normalfile
7 XLS1 xls 35 Normal Success normalfile
8 XLS2 xls 40 Error File damage Damaged
file detection
9 XLS3 xls 44 Normal Success Normalfile
10 DOC1 doc 42 Error Empty Empty
document document
detection
:
50 PPTX3 pptx 85 Normal Success Normalfile

Creating blockchain
First, a blockchain is created. On the first server called chain1, parameter values
of the blockchain are generated in the form of data files and have a set value to
execute the blockchain in the params.dat file. Afterwards, the chain is operated after
checking the initial setting of the params.dat file, and once the blockchain is created
on the node, it will have basic specifications to prepare network activities such as
radio waves and connections. Related information was organized in Table 3.
Connecting blockchain
Next, based on the same settings on the second server after the creation of the
blockchain, the blockchain node can be connected to the first chain1. The first node
created for authorization after initialization is completed, the Wallet Address is

Table 3 Structure of blockchain creation

Basic information
Name Chain1 Version 1.0
Protocol 10,008 Node address [email protected].:xxxx
Permissions Connection, Wallet address 1NnPLDUBauNxqB..38digits
(authority) Transmit, receive, etc. Numeric, English combination
code
Pre-verification of Data in Electronic Trade Blockchain Platform 59

Fig. 2 Structure of blockchain connection

approved, and each node in the blockchain is approved by P2P (Peer to Peer) in
the same way as Fig. 2 between each node, rather than by the server or client.
These configured nodes become peers and form networks. Also, the number of
peers can continue to increase.
Stream Test
Stream is a very important function of the blockchain-based on multi-chain system
currently tested. It is a functional area for documents and searches that are used as
smart contracts for Ethereum, not as asset transactions, where data such as documents
and values that exist inside the blockchain are stored here.
Counterfeit verification test containing electronic document
Electronic document verification tests were conducted by registering docu-
ments directly through web-implemented demonstrations and process components
designed with original reference data. First, the model designed to use the blockchain
system in the fake verification test of electronic documents was tested under the
assumption that the data generated based on the following three items were the orig-
inal: the data to which the original is based shall be defined as the time stamp of
the document, the META information of the document, and the textual information
of the body. The process to construct the verification model is designed as shown
in Fig. 3. The registration process can be targeted at any of the basic archives, or
any newly generated documents, and the registration process is based on the data
generated from the original base data items described above. Electronic Document
A is encrypted through the HASH-256 algorithm, including timestamp, document
META information and body information of the above documents, converted to the
final KEY value to be registered and stored in the blockchain network.
The verification process is shown in Fig. 4 and when the user requests verifica-
tion of Document A-1, which requires comparison and verification, the blockchain
creates the document file information of the document (A-1) for priority comparison
through hash value A-1. Then it performs key verification through KEY1 stored in
the stream (STREAM) inside the blockchain. After that, forgery is confirmed in case
of discrepancy, and the blockchain cannot register new past documents because it
cannot delete or modify past block data that is signed inside.
In order to conduct verification tests based on electronic documents, we have
clearly identified the forged electronic documents in both the case of forging META
information in documents and the case of forging text in the body of documents. In
this process, multi-chain APIs were used.
60 S. Oh et al.

Fig. 3 Registration process of original data

Fig. 4 Process of data verification

Figure 5 also shows the forged META information in the same document. The
first document has META information set to USER, the following documents have
META information set to OTHER, and the text in the body is the same.
Leverage the previously generated stream to store the information from Document
A in the stream. Create a stream and publish Document A to the stream. Stream
registration generates transactions in the blockchain and stores data and key values.
After registration, the original data of document A stored is verified by comparing
meta information with other document B. Register and certify Document B in the
same process.
Through the process, in the event of forgery or tampering with META information
in the original information, the information and document B registered in the chain
Pre-verification of Data in Electronic Trade Blockchain Platform 61

Document A Document B
META DATA : USER META DATA : OTHER

Fig. 5 Test sample that checks the modification of META-DATA in the blockchain

are judged to be different data and will not be finally verified. In addition, verification
was carried out through modification of the text of the same document, specifically,
after setting the META information of the two documents to be verified the same, the
contract amount for Document A was 100 as shown in Fig. 6, the contract amount
for Document B was calculated to be 500, and part of the document contents was
modified. Similarly, the return value was returned to zero and the body text was
identified as modified.
The above verification process was able to determine the integrity of data through
an electronic document forgery experiment applied with blockchain technology and
confirmed that the scenario of blockchain-based electronic document service within
the scope of this study can be used for actual services.
Realization of user based blockchain trade system
In this section, as mentioned in the previous chapter, a pilot system that is readily
available to the general consumer is embodied. The pilot system has established the
verification and smart contract of electronic document files through the blockchain
system so that consumers can directly search and verify them. This can be identified
through Table 4 and Fig. 7. Table 4 is a detailed configuration of the pilot system, and
Fig. 7 shows the front screen on which the pilot system is shown to the consumer.
The first RAW DATA transferred to the blockchain can verify transaction point
information through the Time Verification module at each transaction. In addition,
the first RAW DATA can verify file modifications through a file validation model.
File Viewer is a file converted to SVG to serve consumers, not original data, and
provides point-of-trade certification and file validation marks within the viewer as
62 S. Oh et al.

Document A Document B
Contract amount: 100 Contract amount: 500

Fig. 6 The test sample that applicate modification in text

Table 4 The pilot system that realizes smart contract and data verification
Division Contents
Wallet generating Generate the original address containing user information into wallet
Node information Check connected blockchain network
Repository Create the managing repository for document, data depending on the kind of
asset
Registration Register the data to be created with the original
Verification Compare has for the verification of data authentication based on registered
ledger

shown in Fig. 8. If this is provided through a watermark or QR code, consumers can

check through QR readers.

4 Conclusion

4.1 Summary and Implications of Study

E-document data may be new, but it is already kept by the entity of the transaction—
the agency—that is, it needs to be transferred to the blockchain system for service.
Moreover, it also needs to verify and organize what needs to be provided in order to
Pre-verification of Data in Electronic Trade Blockchain Platform 63

Fig. 7 Blockchain-based electronic transaction system

toolbar

Time of
transacon
cerficaon
mark

Visible through
Visual or QR
readers

File
Verificaon
mark

verificat
ion

Fig. 8 File viewer

provide a platform for the general consumer. For this, filtering of electronic document
files was used to validate electronic documents. Next, a blockchain system was
applied to verify the integrity of the forgery problem that could arise when trading
electronic documents. Finally, we implemented a pilot system that was configured
to be understood by the general consumer when they are provided with electronic
document services. However, there are limitations to the verification of electronic
64 S. Oh et al.

documents conducted in this study, and it is necessary to further study these areas in
the future.

References

1. Zikratov I, Kuzmin A, Akimenko V, Niculichev V, Yalansky L (2017) Ensuring data integrity

using blockchain technology. In: 2017 20th Conference of open innovations association
(FRUCT). IEEE, pp 534–539
2. Swan M (2015) Blockchain: Blueprint for a new economy. O’Reilly Media, Inc.
3. Tang H, Shi Y, Dong P (2019) Public blockchain evaluation using entropy and TOPSIS. Expert
Syst Appl 117:204–210
4. Oh SY, Cho SH, Han SH, Gim GY (2019, July). A study on the pre-verification of data and
the implementation of platform in electronic trade using blockchain. In 2019 20th IEEE/ACIS
International Conference on Software Engineering, Artificial Intelligence, Networking and
Parallel/Distributed Computing (SNPD) pp 320–330, IEEE
5. Tanwar S, Parekh K, Evans R (2020) Blockchain-based electronic healthcare record system
for healthcare 4.0 applications. J Inf Secur Appl 50:102407
6. Lee GY, Kim IH (2019) A study on application of record management system blockchain
technology. Korean J Arch Stud 60:317–358
7. Mao D, Hao Z, Wang F, Li H (2019) Novel automatic food trading system using consortium
blockchain. Arab J Sci Eng 44(4):3439–3455
8. Morris DZ (2016) Leaderless, blockchain-based venture capital fund raises $100 million, and
counting. Fortune (magazine), 05–23
9. Cocco L, Pinna A, Marchesi M (2017) Banking on blockchain: Costs savings thanks to the
blockchain technology. Future Internet 9(3):25
10. McGhin T, Choo KKR, Liu CZ, He D (2019) Blockchain in healthcare applications: research
challenges and opportunities. J Netw Comput Appl 135:62–75
11. Sarmah SS (2018) Understanding blockchain technology. Comput Sci Eng 8(2):23–29
12. Toyoda K, Mathiopoulos PT, Sasase I, Ohtsuki T (2017) A novel blockchain-based product
ownership management system (POMS) for anti-counterfeits in the post supply chain. IEEE
Access 5:17465–17477
13. Saberi S, Kouhizadeh M, Sarkis J, Shen L (2019) Blockchain technology and its relationships
to sustainable supply chain management. Int J Prod Res 57(7):2117–2135
14. Longo F, Nicoletti L, Padovano A, d’Atri G, Forte M (2019). Blockchain-enabled supply
chain: an experimental study. Comput Ind Eng 136:57–69. (Data Source: National Center for
Education Statistics, 1998–2007)
15. Nofer M, Gomber P, Hinz O, Schiereck D (2017) Blockchain. Bus Inf Syst Eng 59(3):183–187
16. Ying W, Jia S, Du W (2018) Digital enablement of blockchain: Evidence from HNA group.
Int J Inf Manage 39:1–4
17. Gordon WJ, Catalini C (2018) Blockchain technology for healthcare: facilitating the transition
to patient-driven interoperability. Comput Struct Biotechnol J 16:224–230
18. Jia B, Zhou T, Li W, Liu Z, Zhang J (2018) A blockchain-based location privacy protection
incentive mechanism in crowd sensing networks. Sensors 18(11):3894
19. Meng W, Tischhauser EW, Wang Q, Wang Y, Han J (2018) When intrusion detection meets
blockchain technology: a review. IEEE Access 6:10179–10188
20. Xu X, Liu Q, Zhang X, Zhang J, Qi L, Dou W (2019) A blockchain-powered crowdsourcing
method with privacy preservation in mobile environment. IEEE Trans Comput Soc Syst
6(6):1407–1419
21. Civelek ME, Özalp A (2018) Blockchain technology and final challenge for paperless foreign
trade. Eurasian Acad Sci Eurasian Bus Econ J 15:1–8
Pre-verification of Data in Electronic Trade Blockchain Platform 65

22. Ghazali O, Saleh OS (2018) A graduation certificate verification model via utilization of the
blockchain technology. J Telecommun Electron Comput Eng (JTEC) 10(3–2):29–34
23. Nizamuddin N, Salah K, Azad MA, Arshad J, Rehman MH (2019) Decentralized document
version control using ethereum blockchain and IPFS. Comput Electr Eng 76:183–197
24. Wang J, Wu P, Wang X, Shou W (2017) The outlook of blockchain technology for construc-
tion engineering management. Front Eng Manage 67–75. (Data Source: National Center for
Education Statistics, 1998–2007)
25. Yang M, Zhu T, Liang K, Zhou W, Deng RH (2019) A blockchain-based location privacy-
preserving crowdsensing system. Future Gener Comput Syst 94:408–418
Study on Security and Privacy
of E-Government Service

Sanghyun Cho, Saeyong Oh, Hogun Rou, and Gwangyong Gim

Abstract In the big data era, a shift from an existing provider-oriented e-government
to a user-centered digital government is required. It is time to implement a consumer-
centered digital government that can proactively provide customized services to the
public rather than services that meet the public’s eye. This study analyzed prior studies
and cases of e-government in order to derive the factors influencing the continuous use
of e-government services provided by central government departments, local govern-
ments and public institutions, and to establish the causal relationship between each
factor. Moreover, based on the information system success model and the technology
acceptance model, the empirical analysis was used to determine how concerns about
quality and privacy of e-government services and security affect usability, user satis-
faction, and willingness to use them continuously. This contributed to presenting the
policy direction and implications to the e-government in the contemporary paradigm
of new IT and the fourth industrial revolution.

Keywords E-government · Quality of e-government services · Security · Privacy ·

Trust · User satisfaction · IS success model · TAM

S. Cho · S. Oh · H. Rou · G. Gim (B)

Department of IT Policy and Management, Soongsil University, Seoul, South Korea
e-mail: [email protected]
S. Cho
e-mail: [email protected]
S. Oh
e-mail: [email protected]
H. Rou
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 67

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_6
68 S. Cho et al.

1 Introduction

With the development of ICT (Information Communication Technology) and the

advent of web browsers, the environment in which anybody can use internet easily
has been created. As a result, people’s Internet access has begun to spread rapidly.
These social changes have brought the implementation of e-government, which uses
ICT to innovate government processes. In particular, the government promoted inno-
vation of administrative services that applied ICT to government affairs to enhance
administrative efficiency and transparency, and drastically improved the quality of
online services to the public, providing the public with easy and convenient access
and usage of various information and services provided by the government, local
governments, public organizations and financial institutions. This has brought new
changes throughout the society as well as in daily life [1].
Accordingly, services that can provide high-quality services and a safe environ-
ment to the level the public wants and meet various demands are now becoming a
necessity, not an option of the e-government. In particular, since the ultimate goal of
e-government is to promote service use and provide safe service rather than to build
a system, it is very important to pinpoint the problems and solutions of users’ needs,
privacy and service safety [1].
Furthermore, it is important for the e-government to seek ways to restore public
trust and continuously develop e-government services through in-depth analysis
of problems and service issues at the service site for those who have used the e-
government service, as it is a very important task for the e-government to raise the
public’s intention to use e-government services through continuous promotion to the
government and provision of various services.
Therefore, this research is meaningful in deriving factors that influence user satis-
faction with e-government services and the intention of continuous use of services
in order to help respond to changes in ICT environment and social structure such as
information technology development, distribution of smartphones, IoT, and artificial
intelligence (AI) in Korea, and implement e-government that can meet the needs
of the people. To accomplish this, the research first analyzed existing research on
e-government to find success factors that can contribute to improving service quality,
and quality (service quality, system quality, information quality). The research hopes
to identify the impact of causal risk (personal information protection, security)
factors on user value (usability, usability, trust) and user satisfaction to study ways
to continuously use e-government services and suggest implications.
Study on Security and Privacy of E-Government Service 69

2 Theoretical Background

2.1 E-Government

The concept of e-government is an extension of the concept that first emerged from
the e-Bank Service mentioned in the 1993 National Performance Review [2], which
was promoted by the Clinton administration in the United States. It refers to the
implementation of a government using information technology to deliver adminis-
trative and public services in a more convenient, customer-oriented and innovative
manner. It also refers to the delivery of necessary administrative and civil services
or administration through electronic means such as the Internet and smartphones.
However, the concept of e-government varies depending on the country pursuing
e-government, academia and consulting institutions working on it, and there is no
agreed definition yet [1].
The concept and definition of e-government are changing little by little, as e-
government can be viewed differently depending on the scope of application and
interpretation from what perspective it interprets as the development of information
and communication technology like Cloud, artificial intelligence(AI) and changes in
the social environment.
It is very important for e-government to build trust with all stakeholders, and if
the people’s trust in e-government increases, they will be able to secure support from
the people. If trust is lowered, the people will be discredited. In this study, among the
factors affecting e-government trust, the focus is on personal information protection
and security, and the factors affecting the service users and the e-government that is
the subject of trust are to be verified with empirical evidence [3].
First, customer orientation is the most important conceptual characteristic in
service quality, one of the factors affecting e-government. The service provided by
e-government should be evaluated by the people and the satisfaction of the service
should be measured from the user’s point of view. Most of the preceding studies on e-
government service quality are based on theoretical background on factors affecting
service acceptance, such as service utilization, user satisfaction, and acceptance from
a consumer-centered perspective. These studies began with the recognition of prob-
lems with relatively low utilization of e-government services despite the high level
of e-government deployment [4].
In an e-government service quality scale study considering relevance, Lee [5]
confirmed, through an empirical analysis, that related quality has a significant impact
on individual efficiency and civic satisfaction in e-government service by adding new
quality to e-government service quality.
70 S. Cho et al.

2.2 Factors for Choosing E-Government

Security in e-government, which provides services based on the Internet environment,

refers to an unauthorized system access, risk and information protection over the
Internet. Due to the importance of information, various studies are being conducted
in the areas of data security and privacy [1].
Prior research on personal information focuses on identifying the causal relation-
ship of factors that affect users’ privacy concerns and intentions to provide personal
information [6], and explaining privacy-related attitudes, intentions and behaviors
centered on users’ risks and benefits from a technical and tool perspective. For e-
government, studies on privacy issues have mainly been addressed in legal and insti-
tutional aspects [7], and prior studies on security in e-government services have
also been mostly case studies that apply security practices to information protection
management systems [1].
In addition, looking at the preceding research on user satisfaction and acceptance
of mobile e-government services, Han identified factors that affect the acceptance of
mobile e-government services and empirically analyzed whether each factor affects
user satisfaction and willingness to reuse, suggesting ways to increase and acti-
vate mobile e-government services [8]. System quality, service quality, information
quality, related quality, and public quality were derived as independent variables, and
user satisfaction and intention to reuse were selected as factors for usability. Usability
and acceptance were parameters to identify causal relationship between each factor.
As a result, system quality, service quality, and related quality among quality factors
have been proven to have positive effects on usability, and information quality has
been proven to have a positive effect on causal relationship to usability [9]. Also,
the casual relationship between the user value (usability, usefulness) factors and
acceptance intention (user satisfaction, intention of reuse) factors have been proven
[10]. This proved that the technology acceptance model and the information system
success model are suitable models to describe mobile e-government services [11].
As proven, e-government is the process of selecting services that provide existing
administrative services in new and innovative ways. The question of what factors
people are considering in choosing new ways of service for the proliferation of e-
government services, activation of use, and search for ways to improve services is a
very important issue [1].

2.3 Information System Success Model (IS Success Model)

This study discussed the impact on the continuous use of e-government services
from the perspective of the information system success model on the reliability of
e-government, users’ satisfaction with the service, and the intention of use due to
risk factors such as service quality, privacy, and security.
Study on Security and Privacy of E-Government Service 71

DeLone and McLean analyzed key research papers related to information system
evaluation to review the requirements and corresponding indicators for information
system performance [12]. The study reviewed system quality, information quality,
usage, satisfaction, individual impact, and organizational performance. Davis also
argued through various literature surveys that perceived usefulness and perceived
use form a causal relationship with the user’s attitude, and that this attitude affects
the position of action and affects actual use [13]. Subsequently, DeLone and McLean
[14] presented a modified model that added a service quality area as a leading quality
factor for user satisfaction. Jin et al. [15] analyzed how mobile payment systems affect
personal benefits based on the information system success model and confirmed that
the quality of the systems and information affects use and user satisfaction, and that
use and user satisfaction are the main factors that have a positive impact on personal
benefits.

2.4 Technology Acceptance Model (TAM)

Technology acceptance model, a model developed based on Davis’s existing rational

theory of behavior, is designed to predict users’ acceptance of new technologies
[15]. It is set to perceived usefulness and perceived ease of use as factors that affect
acceptance, which affect attitudes, affect the intent of action to use them, and explains
the human and human structures in which the intent of action affects actual use [1].
The technology acceptance model is recognized as a concise and very descrip-
tive model in explaining consumers’ acceptance and usage behavior of information
technology. It has been treated as a major model in many empirical studies on the
intent to accept new technologies since the 1990s.
In addition, because the technology acceptance model provides an explanation of
the determinants of computer acceptance and is comprehensive enough to explain
the behavior of users across computer technology and user populations, studies of
what factors influence the adoption of new IT technologies that are emerging with
the advent of the Fourth Industrial Revolution [1].

3 Design of Study

3.1 Research Model

This study will construct the study model by referring to the preceding study and
discuss the results of the analysis. Based on empirically proven variables, quality
factors (service quality, system quality, information quality) and risk factors (person-
al information, security) are set as independent variables. The conclusions are drawn
72 S. Cho et al.

Fig. 1 Research model and hypothesis

based on the research model conducted in advance of establishing user satisfac-

tion and continuous use intent for e-government services by parameters (intentional
usability, perceived usefulness, trust), and dependent variables [1].
Based on these variables, a research model was presented, as shown in Fig. 1,
to demonstrate the causal relationship between the factors, focusing on how quality
and risk factors affect the user’s value factors, how useful and easy to use, and how
users would be satisfied with the service and would intend to continue to use the
e-government service if they were satisfied [1].
The categories of variables being considered in this study are quality factors
(service quality, system quality, information quality) and risk factors (personal
information reporting, security) as independent variables. Recognition factors are
perceived usability, usability, and trust are parameters, and user satisfaction and
continuous use are acceptance factors set as dependent variables. The following
hypotheses were established by reference to the results of the preceding study [1] to
determine the relationship between each variable on how these influencing factors
affect the acceptability factors through the cognitive factors (Table 1).

4 Empirical Analysis

4.1 Data Collection

The survey used in the study was composed of 60 questions, referring to the study
[1]. Each question was measured on a seven-point scale of the recertification, which
measures respondents’ perception of e-government services. IBM SPSS Statistics22
and Smart PLS 2.0 were used as tools for analysis.
Study on Security and Privacy of E-Government Service 73

Table 1 Variable’s operational definition

Factor (measured variable) Operational definition References
Service quality The degree of perceived value of the user for the [1, 16–20]
(SerQ) service
System quality The degree of perceived value of the user to the system
(SysQ)
Information quality The degree of perceived value of the user for the
(InfQ) information
Privacy The degree of perceived value for personal information
(Pri) management, infringement, change, leakage, etc.
Security The degree of perceived value for service safety
(Sec)
Perceived usefulness The degree to which users perceive that they are useful
(PU) for their purpose
Perceived ease of use The degree to which users perceive that the service is
(PEU) easy and convenient
Trust (Tru) The degree to which believe trust the service
User satisfaction The degree of overall user satisfaction after using the
(US) service
Continuous use intention Intended to continue to use the service in the future
(ICU)

First, frequency analysis was performed to examine the demographic characteris-

tics and generalities of the survey respondents. The collected samples were collected
from a total of 318 people using the questionnaire method for those who have used e-
government services. The data were analyzed in 284 parts, excluding those who had
double misses or answered insincerely (34 surveys). In terms of demographic char-
acteristics, 121 males (58.17%) and 87 females (41.83%) were surveyed. In terms
of age distribution, people in their 40s accounted for 34.62%, followed by those in
their 30s (32.69%) and those in their 20s (20.19%).

4.2 Exploratory Factor Analysis

An exploratory factor analysis (EFA) was performed to verify the reliability and
validity of detailed measurement items, and a confirmatory factor analysis (CFA)
was performed to analyze model suitability and verify reliability and validity. The
Cronbach-Alpha coefficient was used to verify reliability. Cronbach-Alpha is an
indication of whether the measured items consist of the same components based
on the mean correlation, and for reliability analysis, Nunally set the reliability of
the concept of the configuration at 0.7 and determined that a value of 0.7 or higher
would not be a problem with the reliability of the target variable [21]. For the purpose
of validation, the degree of association of embedded factors of the variables to be
74 S. Cho et al.

observed was explored in an exploratory factor analysis (EFA) to identify the structure
between factors. A factor analysis indicates that an absolute value of factor loading
is 0.4 or higher is considered a significant variable, and a value greater than 0.5 can
be determined as an important variable [1].
According to the results of an exploratory factor analysis (EFA), reliability anal-
ysis was performed by deleting SysQ2, PU5, PEU 1&5, SerQ 2&3, Pri 4&5, InQ
4&5, and Tru 5. Cronbach-alpha values of all variables were secured above 0.7.

4.3 Confirmatory Factor Analysis

4.3.1 Appropriateness Analysis of Structural Model

The measurement variables extracted through EFA (Exploration Factor Analysis)

were performed using Smart PLS 2.0. Analysis of conformity throughout the struc-
tural model can be verified by indicators such as Redundancy value, R2 value, and
goodness of fit [22].
Looking at the criteria for each metric, the redundancy index is an indicator of the
statistical estimate of the structural model and is evaluated as having the suitability
of the structural model when the redundancy value is positive. Cohen was able to
verify the model’s suitability by the R2 value of the endogenous variables, divided
into ‘high’ if the R2 value was 0.26 or higher, ‘middle’ if the R2 value was greater
than 0.13–0.26 and ‘low’ if they were less than 0.02 and more [23]. Tenenhaus et al.
[24] divided the overall suitability into ‘high’ if the magnitude of the fit is 0.36 or
higher, ‘middle’ if the fit is greater than 0.25 and less than 0.36 and ‘low’ if it is
between 0.1 and 0.25 and less.
In this study, as shown in Table 2, the suitability of the measurement model can
be explanatory because all of the redundancy values have positive values. The R2

Table 2 Reliability analysis of confirmative factor analysis

AVE Composite R2 Cronbach’s Alpha Communality Redundancy
reliability
Pri 0.857 0.947  0.917 0.857 
Sec 0.866 0.970  0.961 0.866 
US 0.782 0.947 0.715 0.930 0.783 0.367
SerQ 0.785 0.916  0.863 0.785 
SysQ 0.741 0.920  0.884 0.741 
Tru 0.850 0.958 0.643 0.941 0.850 0.348
InQ 0.828 0.935  0.896 0.828 
PEU 0.817 0.931 0.616 0.888 0.817 0.251
PU 0.814 0.946 0.677 0.924 0.814 0.197
ICU 0.834 0.962 0.453 0.950 0.834 0.377
Study on Security and Privacy of E-Government Service 75

value is also 0.26 or higher, and the results of the calculation of the adequacy of
the structural model all exceed the criteria and are considered highly suitable. The
overall goodness-of-fit analysis showed that the value was 0.712, indicating a value
greater than 0.36 and higher, indicating a higher model fit.
As a result of checking the suitability of the structural model in this study, as
shown in Table 2, the measurement results for redundancy, R2 , and overall suitability
all exceeded the criteria, indicating that the model is descriptive [1].

4.3.2 Reliability Analysis

The suitability of the previous structural model was shown to be descriptive, and
a Confirmation Factor Analysis (CFA) was conducted to establish the relationship
between the variables. Typically, average variance extractions (AVE: Average Vari-
ance Extracted) and composite reliability (CR: Composite Reliability) are used to
verify reliability. Fornell & Larker stated that reliability is ensured only when the
mean of variance (AVE) extracted on a reliability basis is greater than 0.5 and the
composite reliability (CR) is greater than 0.7. In this study, the mean variance extract
(AVE) of all variables is greater than 0.5 and the composite reliability (CR) is greater
than 0.7 [25]. and all values are above the reference value as shown in Table 2 [1].

4.3.3 Discriminant Validity Analysis

Because the model’s suitability was explained earlier and reliability was ensured
by the positive factor analysis, a discriminant feasibility analysis was conducted to
confirm the validity of the model. Fornell & Larker assesses that the mean variance
extract (AVE) obtained between each potential variable is justified when the square
of the correlation, or square root, appears to be greater than the coefficient of deter-
mination [25]. As analyzed in Table 3 in this study, the highest correlation in the
correlation matrix of the potential variables is that the coefficient of determination
squared by the user satisfaction and confidence (0.801) is 0.642 (0.801800.801), so
the average variance extract (AVE) obtained between all potential variables is higher
than the coefficient of determination. Thus, it can be considered that it is comparable
[1].

4.3.4 Hypothesis Verification

In this study, path analysis was performed to verify each hypothesis using Smart
PLS2.0. The hypothesis verification of the structural model can be verified by the
t-values provided by bootstraping of the PLS. It can be considered statistically signif-
icant at a significant level of 0.01 if the absolute value of the t-values is greater than
1.65, a significant level of 0.05 if greater than 1.96, and a significant level greater
than 2.33, a significant level of 0.02 and 2.58. As a result of the hypothesis test, all
76 S. Cho et al.

Table 3 Result of discriminant validity

Pri Sec US SerQ SysQ Tru InQ PEU PU ICU
Pri 0.857         
Sec 0.798 0.866        
US 0.703 0.765 0.782       
SerQ 0.644 0.637 0.690 0.785      
SysQ 0.694 0.734 0.707 0.732 0.741     
Trus 0.747 0.771 0.801 0.724 0.681 0.850    
InQ 0.677 0.736 0.725 0.709 0.778 0.741 0.828   
PEU 0.657 0.681 0.769 0.697 0.713 0.772 0.728 0.817  
PU 0.692 0.742 0.731 0.711 0.671 0.767 0.743 0.685 0.814 
ICU 0.602 0.645 0.673 0.652 0.584 0.707 0.647 0.562 0.775 0.834

but two of the 14 hypotheses based on the study model had a significant impact and
12 hypotheses were adopted.
Among the two rejected hypotheses, such as the research hypothesis claimed by
Cho et al. [1], the relationship between system quality and perceived usefulness does
not appear to affect the usefulness of e-government services. This is because the
system must be professionally designed for e-government services, and the securing
of system stability (down-error, etc.), and the prompt resolution of problems are
perceived as a natural service that must be performed in the operation of the informa-
tion system. In addition, the relationship between perceived usability and perceived
usefulness is not a useful service for users, although it is easily and conveniently
used at anytime and anywhere. if they do not have the necessary information or are
not helpful to their daily lives.
The results of the path analysis and the adoption of each hypothesis in this study
are shown in Table 4.

5 Conclusions

5.1 Summary of Research Result and Implications

In this study, prior research and case studies on e-government were analyzed to
determine the correlation between the quality factors (service quality, system quality,
information quality), risk factors (personal information protection, security), and user
value factors (understanding usability, perceived usefulness, trust) of e-government
services. Moreover, the study aimed to establish the effect of each factor on user
satisfaction and the degree of continuous use of e-government service. Most of the
hypotheses have been adopted as in previous studies, but the parts that differ from
previous studies are as follows.
Study on Security and Privacy of E-Government Service 77

Table 4 Hypothesis verification result

Hypothesis Path factor T-value Result
Personal information → trust 0.362 3.632*** Adopted
Security → trust 0.482 5.366*** Adopted
User satisfaction → continuous use intention 0.673 14.909*** Adopted
Service quality → perceived easiness 0.275 2.520* Adopted
Service quality → perceived usefulness 0.198 2.421* Adopted
System quality → Perceived easiness 0.244 2.128* Adopted
System quality → perceived usefulness 0.032 0.316 Rejected
Trust → user satisfaction 0.389 3.701*** Adopted
Trust → perceived usefulness 0.365 3.096** Adopted
Information quality → Perceived easiness 0.343 2.434* Adopted
Information quality → perceived usefulness 0.278 2.362* Adopted
Perceived easiness → user satisfaction 0.325 3.961*** Adopted
Perceived easiness → perceived usefulness 0.040 0.382 Rejected
Perceived usefulness → user satisfaction 0.210 2.468* Adopted
*p < 0.05, **p < 0.01, ***p < 0.001

First, System quality does not affect perceived usefulness. E-government ser-vices
should provide convenient access to services that users need anytime, anywhere,
and providing reliable, up-to-date information to service users through continuous
updates is an important factor in determining the usefulness of e-government services.
However, in the case of the system, reflecting the new technology, securing stability
(e.g. down-error), and providing prompt resolution in the event of a problem does
not affect the degree of user convenience. Rather, it is perceived as a natural service
that must be performed in the operation of the information system.
Second, the relationship between perceived usability and perceived usefulness was
found to have no effect on perceived usefulness. Apart from easily and conveniently
using the desired service by accessing the service anytime, anywhere through various
means such as web and mobile, this service can be interpreted as not useful to users
if there is no information they need or if the service and information provided do not
help them in their daily lives.
Based on prior research, this study can be meaningful in that the factors influencing
the continuous acceptability of e-government services were theoretically present-ed.
In addition, the operational definitions, reliability, and validity of the independent
variables, parameters, and dependent variables presented in the study model provided
a theoretical foundation to help empirical research similar to this study in the future.
78 S. Cho et al.

5.2 Limitations of the Study and Direct of Future Study

This paper would like to clarify the limitations of the study that have not been
identified in this research and suggest future research directions to supplement them.
First, the investigation on the perception of e-government services on diverse
subjects was insufficient. The survey targets were limited to those who had experience
using the services provided by the e-government, and no surveys were conducted
on those aged 60 or older. Therefore, a detailed statistical analysis is needed by
expanding the survey respondents’ perception of e-government by those who have
no experience in using e-government services or who do not receive service benefits
due to the digital divide. This will be done by securing many samples.
Second, there was a lack of study on the various factors that affect e-government
trust. Although this study confirmed that trust in e-government is a very important
factor in user satisfaction and continuous use of e-government services, the research
was conducted only in terms of privacy and security. Thus, more detailed research
is needed to present various factors affecting trust in order to accurately investigate
how reliable the e-government is and what factors affect e-government trust.
Finally, since new technologies such as IoT, artificial intelligence (AI), and block
chain should be applied and policies and laws should be supported to implement
intelligent e-government suitable for the era of the Fourth Industrial Revolution, it
is necessary to study what kind of perceptions that users who use e-government ser-
vices have an awareness of e-government-related systems and how these perceptions
affect the intention of continuous use of e-government services.

References

1. Cho SH, Oh SY, Rou HG, Gim GY (2019) A study on the factors affecting the contin-
uous use of e-government services-focused on privacy and security concerns. In: 2019
20th IEEE/ACIS international conference on software engineering, artificial intelligence,
networking and parallel/distributed computing (SNPD). IEEE, pp 351–361
2. National Intelligence Service, Ministry of Public Administration and Security, Ministry of
Knowledge Economy, & Korea Communications Commission (2009) National Information
Protection White Paper. Seoul, Korea
3. Aaker D (2003) The power of the branded differentiator. MIT Sloan Manage Rev 45(1):83;
Al-Omari H, Al-Omari A (2006) Building an e-Government e-trust infrastructure. Am J Appl
Sci 3(11):2122–2130
4. Kwon YM (2014) A study on the effect of service quality on service satisfaction and govern-
ment trust in the government 3.0 era—focusing on the information gap between users in
the metropolitan area. Doctoral dissertation. Available from RISS Dissertation and Theses
database. (No. 13417330)
5. Lee CE (2008) The effect of relationship quality on citizen satisfaction with electronic govern-
ment services. Doctoral dissertation. Available from RISS Dissertation and Theses database.
(No. 11274561)
6. Lee HG, Lee SH (2009) Analysis of prior studies for elucidating future directions of information
privacy: focusing on information privacy concerns online. Informatization Policy 16(2):3–16
Study on Security and Privacy of E-Government Service 79

7. Choi BM, Park MJ, Chae SM (2015) A study on factors affecting personal information
technology acceptance behavior. Inf Syst Rev 17(3):77–94
8. Han KH (2012) An empirical study on the influencing factors of intention to adoption of
mobile e-government service. Doctoral dissertation. Available from RISS Dissertation and
Theses database. (No. 12683689)
9. Elliman T (2006) Generating citizen trust in e-government using a trust verification agent: a
research note. In: European and mediterranean conference on information systems
10. Warren SD, Brandeis LD (1890) The right to privacy. Harvard Law Rev 193–220
11. Kim BS, Lee J, Kim KK (2006) The impact of perceived trust, risk, usability, and conve-
nience on intention to use e-government service: focused on online application and service.
Informatization Policy 13(4):186–202
12. DeLone WH, McLean ER (1992) Information systems success: the quest for the dependent
variable. Inf Syst Res 3(1):60–95
13. Davis FD (1989) Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Q 319–340
14. Delone WH, McLean ER (2003) The DeLone and McLean model of information systems
success: a ten-year update. J Manage Inf Syst 19(4):9–30
15. Jin JS, Kim HM, Park JS (2017) A study on behavior in using fin-tech based on life style types.
J Inf Technol Serv 16(1):119–138
16. Pitt LF, Watson RT, Kavan CB (1995) Service quality: a measure of information systems
effectiveness. MIS Q 173–187
17. In KY (2016) A study of the effect of service quality to the continuous usage of e-government
services. Master thesis. Available from RISS Dissertation and Theses database. (No. 14059381)
18. Davis FD (1985) A technology acceptance model for empirically testing new end-user
information systems: theory and results. Doctoral dissertation, Massachusetts Institute of
Technology
19. Kim BS (2018) An empirical study on factors affecting the operational performance of the
war game system of the ROK air force: focused on theater-level exercise model. Doctoral
dissertation. Available from RISS Dissertation and Theses database. (No. 14730384)
20. Weinstein L, Neumann PG (2000) Internet risks. Commun ACM 43(5):144–144
21. Nunnally JC (1978) Psychometric theory. McGraw-Hill, New York
22. Chin WW (1998b) Issues and opinion on structural equation modeling. Mis Q 22(1)
23. Cohen J (1988) Statistical power analysis for the behavioral sciences. Psychology Press
24. Tenenhaus M, Vinzi VE, Chatelin YM, Lauro C (2005) PLS path modeling. Comput Stat Data
Anal 48(1):159–205
25. Fornell C, Larcker DF (1981) Evaluating structural equation models with unobservable
variables and measurement error. J Mark Res 18(1):39–50
Applying Hofstede’s Culture Theory
in the Comparison Between Vietnam
and Korean E-Government Adoption

Hung-Trong Van, Simon Gim, Euntaek Lim, and Thi-Thanh-Thao Vo

Abstract Due to the rapid development of information technology (IT), E-

government has become a popular issue all over the world. E-government system
has also become the standard to comparing the technical development background of
many countries around the world. Keeping up with the trend, Vietnam Government
has been focusing on developing E-Government system for years and E-government
has become one of the most important priorities. Compared to Vietnam, which is
at the beginning stage, Korean E-Government is one of the top systems in the E-
Government World Ranking. Based on Hofstede’s National Culture theory, this study
finds the difference of E-government system/service between Vietnam and Korea
under the evaluations of citizens.

Keywords E-government · Korea · Vietnam · Difference · National culture

theory · TAM · IS success model

1 Introduction

Many policy makers and politicians in both developing and developed countries
have been attracted to use E-government as a way to develop their countries. Unlike

H.-T. Van (B) · T.-T.-T. Vo

Faculty of Digital Economy & E-Commerce, Vietnam–Korea University of Information and
Communication Technology, Danang, Vietnam
e-mail: [email protected]
T.-T.-T. Vo
e-mail: [email protected]
S. Gim
SNS Marketing Research Institute, Soongsil University, Seoul, Korea
e-mail: [email protected]
E. Lim
Graduate School of Business Administration, Soongsil University, Seoul, Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 81

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_7
82 H.-T. Van et al.

other systems which are still in controversy regarding privatization and globalization
[1], E-government is recognized as an effective solution which has been quickly
developing and spreading across the world. To establish this E-government, many
countries have been investing a lot of money and political commitment [1]. In many
developing countries such as Indonesia, India, and Vietnam, E-government initiatives
have been collected to create the opportunities for ICT and economic development
[2]. E-government system is viewed as an effective tool to reduce corruption and
enhance economic development. Not only that, it can improve the services’ quality
for their citizens [3].
Although E-government systems have been used in both developing and devel-
oped countries, not all E-government system’s diffusion has gained expected indi-
cated outcomes [4]. In case of developing countries, the diffusion process remains
slow due to the incessant shortage of resources [5]. Almost all the developing coun-
tries suffer from lack of finance due to weak national information infrastructure and
insufficient knowledge or skill to develop effective and appropriate strategies for
E-government. This research focuses on finding the factors affecting the intention
to use E-government by comparing Korea and Vietnam citizens’ behavior based on
TAM and IS success Model. Furthermore, the Hofstede’s theory on national culture
is used to find out the difference between Korea and Vietnam.

2 Literature Review

2.1 E-Government in Korea

Korean E-government was first implemented in 1978 with ambition to become the
E-government with the highest developing E-government index in the world. To
start with, significant policies and strategies were established by the Korean govern-
ment to promote the growth of government service under the aid of ICT. After
40 years of implementing E-government, there has been many noticeable remarks
made in Korea. First, government’s agencies and departments were computerized
and connected to the Internet in the late 1970s. Secondly, several important projects
were implemented and deployed including the Five National Computer Network and
Comprehensive Plan for Korea Information Infrastructure Establishment. In 2001, E-
government development strategies which concentrated on contributions and diver-
sification of delivery channel in public services were established [2, 6]. Through
numbers of concentrated projects, Korean government built a high-speed commu-
nication network and proper storing systems to put government record in digital
formats. These formats provided solid foundation for E-government implementation
in Korea [6].
Applying Hofstede’s Culture Theory in the Comparison … 83

2.2 E-Government in Vietnam

Vietnam government has been providing E-government services to citizens for many
years [2]. In most developing countries, E-government is recognized as a potential
tool to improve government institutions’ capacities and offer chances to better solve
problems in the public administration [7]. Nevertheless, as a developing country with
low level of IT adoption and economic development, Vietnam has numerous issues
in applying and implementing E-government services/system. Therefore, the current
statement of Vietnam E-government is still far below the government’s expectation.
It does not attract significant concern from citizens as well as business areas [8].
E-government services are still in its inception with citizens and businesses, and the
outcome of E-citizenship dimension is even more limited [9].

2.3 Information System’s Theories

Technology Acceptance Model (TAM) of Davis and the original and updated IS
success models by DeLone and McLean are considered as one of the most reputable
models that have been extensively used for predicting and explaining underlying
factors. TAM theorizes that the behavioral intentions of one person are decided by
two belief constructs as perceived usefulness and perceived ease of use. The aims of
IS success model are defining and clarifying the relationships among six of the most
important dimensions which evaluate any information systems.

2.4 Hofstede’s National Culture Theory

National culture theory of Hofstede is a well-known framework for any cross-cultural

research. In this theory, the effects of national culture on many research areas are
described based on the values of members in that society and how the value links
with human behavior. This is done by using a structure which is acquired from factor
data analysis [2]. According to Hofstede, the differences of national cultures are
identified based on six primary dimensions. These dimensions are power distance,
collectivism and individualism, uncertainty avoidance, masculinity and femininity,
long-term orientation and short-term orientation, and indulgence and restraint.

3 Research Model and Hypotheses

Considering the IS success model, TAM, national culture theory, and other researches
about E-government, the research model was proposed in Fig. 1 [2].
84 H.-T. Van et al.

Information Quality
Quality

Service Quality Perceived Ease of Use

System Quality

Intention to Use

Social Influence
Environment

Awareness Perceived Usefulness

Government Regulation
Culture Effect

Power Distance Uncertainty Avoidance Collectivism/Individualism

Masculinity/Femininity Indulgence/Restraint Long/Short Term Orientation

Fig. 1 Research model

According to Delone and McLean, information quality is identified with how

much the given information fits the requirements of client [10]. In E-government
literature, information quality is characterized with respect to some subsets or some
assortments of the accompanying traits: relevancy, validity (accuracy), complete-
ness (breadth and comprehensiveness), understandability, accessibility, reliability
(dependability), timeliness (currency), concise presentation, authority, objectivity,
security, appropriate amount (quantity), perceived value and freedom from error [2].
System quality is characterized as the information system’s quality processing that
incorporates data and software parts. In the context of E-government services, govern-
ment sites ought to give prompt, simple, and easy information approach regarding
technology [11]. This gives a definite influence on perceived usefulness and perceived
ease of use because of improved adequacy and productivity [12]. Service quality is
related to the degree to which the services are delivered to address customers’ needs
and in relation to the entity managing the support of the system [10], particularly
in the context of E-government, service quality plays an important role [12]. More-
over, users in every service encounter normally expect a satisfactory level of service
quality when engaging in any E-government services [12].
In accordance with Venkatesh and Davis, social influence is one of the key
elements in discovering usage behavior as well as users’ acceptance of any novel
Applying Hofstede’s Culture Theory in the Comparison … 85

IT [13]. Taylor and Todd pointed out that social influence refers to the subjec-
tive norms such as peers influence, superior influence as well as the opinion from
other people [14]. Additionally, the approving or disapproving perceptions of people
towards the E-government usage might be affected by the perceptions of their family,
colleagues, or acquaintances influence [2]. Awareness is related to an understanding
of others’ behaviors, from that to provide a framework for your own behavior [15].
In E-government context, using mass media and carrying out workshops are one
of methods for introducing the concept as well as encourage citizens to use E-
government services in their daily life. Government regulation refers to government
laws, regulations, and policies that any associations must comply with it. Regulation
environment is considered as important factor in IT innovation adoption [16]. In the
context of E-government, it is necessary to have new policies and regulations for
developing E-government assimilation. If government doesn’t offer a suitable regu-
latory framework for using E-government, associations might discourage or corrupt
in using it [2].
Perceived usefulness refers to the assessment of users regarding task-oriented
outcomes. It includes information regarding the extrinsic of the task such as how IT
could help to enhance task efficiency and effectiveness [17]. Perceived usefulness is
seen as a significant factor in determining the adoption of innovations [18]. According
to Kumar, the adoption of E-government is affected directly by the extend of perceived
the usefulness [19]. Perceived ease of use is defined as the extent to which one person
has the belief that it would be effortless in using a particular system [20].
Power distance refers to the extent to which the unequal power distribution in
organizations is accepted by the less powerful members. People have inclination to
accept and finish their tasks as well as duties which is assigned to them by their
bosses [21]. In high power distance society, E-government activities as well as IT
adoption could be perceived as a threat to the hierarchy [22]. According to Geert and
Jan, uncertainty avoidance is related to how much obscure or dubious circumstances
danger individuals in a society [23]. In simple way, uncertainty avoidance is the extent
that one society’s citizens feel intimidated and try to avoid dubious circumstances
by ignoring abnormal ideas and behaviors, and designing formal regulations and
policies [24]. In high uncertainty avoidance society, citizens usually feel threaten by
unstructured, uncertain, and ambiguity circumstances [25]. Collectivism is related to
a society in that the connection among people are close; everybody has belief to take
care themselves as well as people around them [26]. In individualism cultures, people
believe in their decisions and have more self-oriented characteristics. Masculinity,
in contrast to femininity, is defined as a culture where there is a clear separation
between the emotional roles between the two genders. According to Hofstede and
his partners, a higher masculinity society performs a competitive, productive, and
assertive centered culture [26]. Conversely, femininity culture is related to a more
solidarity, equality, consensus seeking social relationships, and concerning centered
culture [27]. In high masculinity society, people are more interested in the useful-
ness of given technology regardless of whether it is easy or difficult to use [28].
Long-term orientation culture is differentiated by values [26]. Conversely, short-
term orientation culture has belief in quick outcomes and relies upon regard for the
86 H.-T. Van et al.

past focusing on steadiness and custom [29]. In long term orientation society, the
connected values are persistence and carefulness. Meanwhile, the associated values
in short-term orientation culture are fulfilment with social commitments and respect
to customs [2]. Indulgence is identified with a culture that moderately free delight
of elemental and natural human ambitions associated with having good times and
appreciating life are permitted [26]. Moreover, restraint refers to a culture that the
needs’ gratification is controlled and directed by severe social standards’ ways [26].
Society with high indulgence index does not apply strict social norms to the fulfil-
ment of people about basic needs and desires. Accordingly, people can utilize new
technologies when they want it. Conversely, in restraint society, individuals have the
constraint on the adoption with new innovations, adopting even their life [26]. The
hypotheses are proposed:
H1: Information Quality (InfQ) has positive influence on Perceived Usefulness
(PU)
H2: Information Quality has positive influence on Perceived Ease of Use (PEU)
H3: System Quality (SysQ) has positive influence on Perceived Usefulness
H4: System Quality has positive influence on Perceived Ease of Use
H5: Service Quality (SerQ) has positive influence on Perceived Usefulness
H6: Service Quality has positive influence on Perceived Ease of Use
H7: Social Influence (SI) has positive influence on Perceived Usefulness
H8: Awareness (AW) has positive influence on Perceived Usefulness
H9: Government Regulation (GR) has positive influence on Perceived Usefulness
H10: Perceived Ease of Use has positive influence on the Perceived Usefulness
H11: Perceived Ease of Use has positive influence on the Intention to Use E-
government (In)
H12: Perceived usefulness has positive influence on the Intention to use
E-government
H13: Power Distance (PD) moderates the relationship between Perceived ease of
use and Intention to use E-government
H14: Power Distance moderates the relationship between Perceived usefulness
and Intention to use E-government
H15: Uncertainty Avoidance (UA) moderates the relationship between Perceived
ease of use and Intention to use E-government
H16: Uncertainty Avoidance moderates the relationship between Perceived
usefulness and Intention to use E-government
H17: Collectivism/Individualism (CO) moderates the relationship between
Perceived ease of use and Intention to use E-government
H18: Collectivism/Individualism moderates the relationship between Perceived
usefulness and Intention to use E-government
H19: Masculinity/Femininity (MA) moderates the relationship between Perceived
ease of use and Intention to use E-government
H20: Masculinity/Femininity moderates the relationship between Perceived
usefulness and Intention to use E-government
Applying Hofstede’s Culture Theory in the Comparison … 87

H21: Long Term Orientation/Short Term Orientation (LT) moderates the relation-
ship between Perceived ease of use and Intention to use E-government
H22: Long Term Orientation/Short Term Orientation moderates the relationship
between Perceived usefulness and Intention to use E-government
H23: Indulgence/Restraint (ID) moderates the relationship between Perceived
ease of use and Intention to use E-government
H24: Indulgence/Restraint moderates the relationship between Perceived useful-
ness and Intention to use E-government

4 Data Analysis

The research model was evaluated by using a designed questionnaire to collect

data. The instruments were measured based on 7-point Likers-scales from strongly
disagree (1) to strongly agree (7). The survey was carried out for 30 days from
September 2015. 462 appropriate responses out of 480 returned responses were used
for data analysis giving 96.25% of response rate.
Table 1 shows the results for convergent validity testing including AVE and CR.
If AVE is greater than 0.50 and CR greater than 0.7 and CR is over AVE of the total
variance, convergent validity is established [2] and the convergent validities of all 9
factors were confirmed.
In Table 2, it can be seen that service quality does not have effect to the perceived
usefulness of E-government and perceived ease of use does not affect to the intention
to use E-government with the p-value higher than 0.1 [2].
According to Table 3, in case of Korea E-government, the relationship between
service quality and perceived usefulness as well as the relationship between system
quality and perceived usefulness are rejected with p-values higher than 0.1 [2].
According to Table 4, in case of Vietnam E-government, government regulation
and awareness do not have effect on the perceived usefulness of E-government with
p-values higher than 0.1 [2].
In Table 5, it shows that power distance cannot moderate the relationship between
perceived ease of use and intention to use in the comparison between Korea and
Vietnam because the p-value are higher than 0.1. However, power distance can
moderate the relationship between perceived usefulness and intention to use E-
government in the comparison between Korea and Vietnam when the p-value is
less than 0.01 [2]. Uncertainty avoidance cannot moderate the relationship between
perceived ease of use and intention to use as well as perceived usefulness and intention
to use E-government in the comparison between Korea and Vietnam with p-values

Table 1 CR and AVE

AW GR PU SI InfQ SysQ PEU SerQ In
CR 0.865 0.92 0.88 0.896 0.917 0.908 0.917 0.799 0.899
AVE 0.682 0.741 0.709 0.684 0.735 0.711 0.786 0.572 0.748
88 H.-T. Van et al.

Table 2 Structural paths assessment and hypothesis test

Estimate S.E. C.R. p Label
PEU <--- InfQ 0.307 0.060 40.824 *** Supported
PEU <--- SysQ 0.418 0.060 60.481 *** Supported
PEU <--- SerQ 0.122 0.042 20.721 0.007 Supported
PU <--- PEU 0.329 0.053 60.234 *** Supported
PU <--- GR 0.234 0.044 40.864 *** Supported
PU <--- SI 0.136 0.035 30.410 *** Supported
PU <--- InfQ 0.265 0.057 40.401 *** Supported
PU <--- SysQ 0.109 0.058 10.790 0.073 Supported
PU <--- SerQ −0.054 0.036 −10.415 0.157 Rejected
PU <--- AW 0.176 0.037 40.199 *** Supported
In <--- PEU 0.082 0.057 10.449 0.147 Rejected
In <--- PU 0.716 0.068 100.404 *** Supported

Table 3 Korea structural paths assessment test

Estimate S.E. C.R. p Results
PEU <--- InfQ 0.563 0.103 5.411 *** Supported
PEU <--- SysQ 0.207 0.087 2.082 0.037 Supported
PEU <--- SerQ 0.156 0.059 2.382 0.017 Supported
PU <--- PEU 0.2 0.083 2.423 0.015 Supported
PU <--- GR 0.377 0.083 4.024 *** Supported
PU <--- SI 0.168 0.048 3.281 0.001 Supported
PU <--- InfQ 0.418 0.112 3.703 *** Supported
PU <--- SysQ −0.088 0.085 −0.91 0.363 Rejected
PU <--- SerQ −0.01 0.056 −0.163 0.871 Rejected
PU <--- AW 0.122 0.056 2.191 0.028 Supported
In <--- PEU 0.213 0.078 2.444 0.015 Supported
In <--- PU 0.592 0.09 5.884 *** Supported

higher than 0.1 [2]. Collectivism/individualism cannot moderate the relationship

between perceived ease of use to intention to use and perceived usefulness to inten-
tion to use E-government in the comparison between Korea and Vietnam because
p-values are higher than 0.1 [2]. Masculinity/femininity cannot moderate the rela-
tionship between perceived ease of use and intention to use as well as perceived
usefulness and intention to use E-government in the comparison between Korea and
Vietnam because with p-values higher than 0.1 [2]. Long term orientation/short term
orientation cannot moderate the relationship between perceived ease of use and inten-
tion to use as well as perceived usefulness and intention to use E-government in the
comparison between Korea and Vietnam because p-values are higher than 0.1 [2].
Applying Hofstede’s Culture Theory in the Comparison … 89

Table 4 Vietnam structural paths assessment test

Estimate S.E. C.R. p Results
PEU <--- InfQ 0.131 0.064 1.69 0.091 Supported
PEU <--- SysQ 0.547 0.083 6.187 *** Supported
PEU <--- SerQ 0.139 0.056 2.114 0.035 Supported
PU <--- PEU 0.322 0.066 4.891 *** Supported
PU <--- GR −0.058 0.072 −0.848 0.396 Rejected
PU <--- SI 0.177 0.051 2.695 0.007 Supported
PU <--- InfQ 0.146 0.063 1.914 0.056 Supported
PU <--- SysQ 0.39 0.078 4.693 *** Supported
PU <--- SerQ 0.13 0.048 2.315 0.021 Supported
PU <--- AW 0.088 0.081 1.281 0.2 Rejected
In <--- PEU 0.154 0.079 2.053 0.04 Supported
In <--- PU 0.585 0.091 6.73 *** Supported

Indulgence/restraint cannot moderate the relationship between perceived usefulness

and intention to use E-government in the comparison between Korea and Vietnam
because p-value is higher than 0.1. However, indulgence can be the moderator vari-
able of the relationship between perceived ease of use and intention to use in the
comparison between Korea and Vietnam because p-value is 0.017 (<0.1) [2].
Table 6 shows that power distance does not affect the relationship between
perceived usefulness and intention to use E-government for Korean (p-value =
0.202). Diversely, when p-value is less than 0.1, the power distance has negative effect
on the relationship between perceived usefulness and intention to use E-government
for Vietnamese (Beta = −0.133, p-value = 0.074) [2]. Indulgence/restraint has posi-
tive effect on the relationship between perceived ease of use and intention to use
E-government for Korean (Beta = 0.045, p-value = 0.085). Diversely, when p-value
is more than 0.1, indulgence/restraint does not have effect on the relationship between
perceived ease of use and intention to use E-government for Vietnamese (p-value =
0.864) [2].

5 Conclusion

Based on the result, in case of the main model, the perceived usefulness factor is the
strongest factor in both in Vietnam and Korea (0.716). This indicates that citizens
will use E-government system/services when they get the benefits. After that, the
perceived ease of use factor also has strong effect on the perceived usefulness factor.
It means that when citizens feel any ICT system is easy to use, they will get their
usefulness. Besides, the quality group with three factors which are services quality,
information quality, and system quality also have strong effect to the perceived ease
90 H.-T. Van et al.

Table 5 Checking moderator factor

Power distance
Model NPAR CMIN DF p CMIN/DF
255 3893.102 1956 0.000 1.990
Estimate S.E. C.R. p Results
In <--- PEUxPD 0.031 0.024 1.312 0.190 Rejected
In <--- PUxPD −0.109 0.031 −3.466 *** Supported
Uncertainty avoidance
Model NPAR CMIN DF p CMIN/DF
255 4587.417 1956 0.000 2.345
Estimate S.E. C.R. p Result
In <--- PEUxUA 0.029 0.035 0.851 0.395 Rejected
In <--- PUxUA −0.016 0.032 −0.509 0.611 Rejected
Collectivism/individualism
Model NPAR CMIN DF p CMIN/DF
255 4444.614 1956 0.000 2.274
Estimate S.E. C.R. p
In <--- PEUxCO 0.035 0.032 1.114 0.265 Rejected
In <--- PUxCO 0.044 0.027 1.621 0.105 Rejected
Masculinity/femininity
Model NPAR CMIN DF p CMIN/DF
255 4027.732 1956 0.000 2.059
Estimate S.E. C.R. p Result
In <--- PEUxMA 0.021 0.024 0.890 0.374 Rejected
In <--- PUxMA −0.065 0.041 −1.578 0.115 Rejected
Long term orientation/short term orientation
Model NPAR CMIN DF p CMIN/DF
255 4350.724 1956 0.000 2.224
Estimate S.E. C.R. p Result
In <--- PEUxLT 0.039 0.031 1.239 0.215 Rejected
In <--- PUxLT 0.036 0.027 1.337 0.181 Rejected
Indulgence/restraint
Model NPAR CMIN DF p CMIN/DF
255 4046.794 1956 0.000 2.069
Estimate S.E. C.R. p Result
In <--- PEUxID 0.059 0.025 2.381 0.017 Supported
In <--- PUxID −0.027 0.025 −1.084 0.278 Rejected
Applying Hofstede’s Culture Theory in the Comparison … 91

Table 6 Checking national culture dimensions difference between Korea and Vietnam
Korea Vietnam
Power distance Indulgence/restraint Power distance Indulgence/restraint
In <--- PU 00.555 (p = 00.638 (p =
***) ***)
In <--- PD 00.006 (p = 00.006 (p =
0.865) 0.912)
In <--- PUxPD −00.042 (p = −00.133 (p =
0.202) 0.074)
In <--- PEU 00.215 (p = 0.007) 00.282 (p = 0.001)
In <--- ID 00.018 (p = 0.588) −00.038 (p = 0.645)
In <--- PEUxID 00.045 (p = 0.085) −00.010 (p = 0.864)

of use with high value which are 0.122, 0.307, 0.418 respectively. Comparing the
factors in quality group, system quality is the strongest which indicates that the role
of the system is continuous and safe [2]. On the other hand, information quality also
has the important role which affect both Perceived Usefulness and Perceived Ease of
Use with the value of 0.265 and 0.307 respectively. In this research, in Environment
group, the government regulation is the strongest factor which has direct affect to the
perceived usefulness. Regulation is very important for the government because with
the good regulation, government can supervise their state officials and their citizens.
Besides, awareness and influence statement have the similar value with 0.176 and
0.136. For the E-government, awareness factor explains how people can understand
the ICT services. Similarly, social influence statement strongly affects the perceived
usefulness. If some people do not know about online services, the introduction from
other people can help those citizens [30]. Regarding the comparison between Vietnam
and Korea, there are many differences. For environment group, in Vietnam, the
government still does not have good regulations to implement the E-government
services/system. People do not have good condition to apply and use the new services
[2]. Besides, the research shows that people in Vietnam can get the benefit and
information of the E-government services/system form the others because the social
influence factor which is one of the important factors (p-value = 0.007).
Based on the National Culture Theory of Hofstede, which applies the main
model, this research cannot find the moderation of long/short term orientation, uncer-
tainty avoidance, and collectivism/individualism from the perceived usefulness and
perceived ease of use to intention to use the E-government (p-value > 0.1). Moreover,
indulgence/restraint only moderates the effect of perceived ease of use to intention to
use and power distance only moderate the effect of perceived usefulness to intention to
use E-government. Furthermore, In Vietnam, power distance factor only has negative
effect to the relationship between perceived usefulness and intention to use. Mean-
while, it has no effect in Korea. The country with high power distance would have
negative attitude for using and implementing E-government services [31]. Vietnam
is the country with high power distance, people can affect the behavior intention
92 H.-T. Van et al.

to use in many ICT services [2]. The problem to implement the E-government in
Vietnam also comes from their culture. Diversely, indulgence/restraint has positive
effect to the interaction of perceived ease of use and intention to use E-government
in Korea (p-value = 0.085). However, it has no effect in Vietnam. In the countries
which have indulgence more than restraint, the ICT service quality is more relevant
[32]. It is the reason why the public services and E-government services in Korea
has the high ranking in the world.

References

1. Stoltzfus K (2005) Motivations for implementing E-government: an investigation of the global

phenomenon. In: Proceedings of the 2005 national conference on digital government
2. Van HT, Kim B, Lee SY, Gim GY (2019) The difference of intention to use E-government
based on national culture between Vietnam and Korea. In: 2019 20th IEEE/ACIS international
conference on software engineering, artificial intelligence, networking and parallel/distributed
computing (SNPD): 409–420
3. Yadav V (2011) Political parties, business groups, and corruption in developing countries.
Oxford University Press, Oxford
4. Ferro E, Sorrentino M (2010) Can intermunicipal collaboration help the diffusion of
E-government in peripheral areas? Evidence from Italy. Gov Inf Q 27(1):17–25
5. Lin F, Fofanah SS, Liang D (2011) Assessing citizen adoption of E-government initiatives in
Gambia: a validation of the technology acceptance model in information systems success. Gov
Inf Q 28(2):271–279
6. Muguti KP (2013) Mobile government adoption in Zimbabwe: lessons learnt from South Korea
and South Africa. Global Information Telecommunications & Technology Program, School of
Innovation
7. Schuppan T (2009) E-government in developing countries: experiences from sub-Saharan
Africa. Gov Inf Q 26(1):118–127
8. Khuong MV, West DM (2005) E-government and business competitiveness: a policy review.
VNCI Vietnam Competitiveness Initiative
9. Tuan ND (2007) E-government in Vietnam: an assessment of province websites (Master thesis).
Gadjah Mada University, Indonesia
10. Delone WH, McLean ER (2003) The DeLone and McLean model of information systems
success: a ten-year update. J Manag Inf Syst 19(4):9–30
11. Schaupp LC, Carter L, McBride ME (2010) E-file adoption: a study of US taxpayers’ intentions.
Comput Hum Behav 26(4):636–644
12. Floropoulos J, Spathis C, Halvatzis D, Tsipouridou M (2010) Measuring the success of the
Greek taxation information system. Int J Inf Manag 30(1):47–56
13. Venkatesh V, Davis FD (2000) A theoretical extension of the technology acceptance model:
four longitudinal field studies. Manag Sci 46(2):186–204
14. Taylor S, Todd PA (1995) Understanding information technology usage: a test of competing
models. Inf Syst Res 6(2):144–176
15. Dourish P, Bellotti V (1992) Awareness and coordination in shared workspaces. In: Proceedings
of the 1992 ACM conference on computer-supported cooperative work, pp 107–114
16. Hart P, Saunders C (1997) Power and trust: critical factors in the adoption and use of electronic
data interchange. Organ Sci 8(1):23–42
17. Gefen D (2000) E-commerce: the role of familiarity and trust. Omega 28(6):725–737
18. Tan M, Teo TS (2000) Factors influencing the adoption of Internet banking. J Assoc Inf Syst
1(1):5
Applying Hofstede’s Culture Theory in the Comparison … 93

19. Kumar V, Mukerji B, Butt I, Persaud A (2007) Factors for successful E-government adoption:
a conceptual framework. Electron J E-gov 5(1)
20. Davis FD (1989) Perceived usefulness, perceived ease of use, and user acceptance of
information technology. MIS Q 13:319–340
21. Hofstede G (1984) Culture’s consequences: international differences in work-related values,
vol 5. Sage, London
22. Islam N (2004) Sifarish, sycophants, power and collectivism: administrative culture in Pakistan.
Int Rev Admin Sci 70(2):311–330
23. Geert H, Jan HG (1991) Cultures and organizations: software of the mind. McGrawHill, USA
24. McCoy S, Galletta DF, King WR (2005) Integrating national culture into IS research: the need
for current individual level measures. Commun Assoc Inf Syst 15(1):12
25. Ford DP, Connelly CE, Meister DB (2003) Information systems research and Hofstede’s
culture’s consequences: an uneasy and incomplete partnership. IEEE Trans Eng Manag
50(1):8–25
26. Hofstede GJ, Minkov M (2010) Cultures and organizations: software of the mind. Revised and
expanded, 3rd edn. McGraw-Hill, New York
27. Erumban AA, DeJong SB (2006) Cross-country differences in ICT adoption: a consequence
of culture? J World Bus 41(4):302–314
28. Srite M (1999) The influence of national culture on the acceptance and use of information
technologies: an empirical study. In: AMCIS 1999 proceedings, vol 355
29. Bouaziz F (2008) Public administration presence on the web: a cultural explanation. Electron
J E-gov 6(1)
30. Al-Somali SA, Gholami R, Clegg B (2009) An investigation into the acceptance of online
banking in Saudi Arabia. Technovation 29(2):130–141
31. Aida A, Majdi M (2014) National culture and E-government services adoption Tunisian case.
Int J Bus Econ Strategy (IJBES) 1:4
32. Gracia DB, Ariño LVC, Blasco MG (2015) The effect of culture in forming e-loyalty intentions:
a cross-cultural analysis between Argentina and Spain. BRQ Bus Res Q 18(4):275–292
Study on Business Strategy
Quantification Using Topic Modeling
and Word Embedding: Focusing
on ‘Virtual Reality’ and ‘Augmented
Reality’

Siyoung Lee, Sungwoong Seo, Hyunjae Yoo, and Gwangyong Gim

Abstract This study analyzed articles related to “Virtual Reality” and “Augmented
Reality,” the main elements of the Fourth Industrial Revolution, and reviewed the
views that the media has by using text mining techniques. For this, 14,443 news arti-
cles on “Virtual Reality” and “Augment Reality” from 2011 to 2020 were collected
from the BigKinds of Korea Press Foundation. For the analysis of collected datasets,
the LDA algorithm-based topical modeling techniques implemented in Python3
languages were used. The number of topics suitable for the collected dataset using
coherence was determined to be 14. The calculated topic was given the title of the
topic through expert survey, and word vector was created using Word2vec, one of
the word-embedding techniques. This word vector was then used to quantify the
analyzed topic. Through this, quantitative expressions of strategic perspectives were
made and changes in topics around the year 2016 were visualized when the issue of
the “fourth industrial revolution” began to be discussed.

Keywords Latent Dirichlet allocation · Topic modeling · Word2vec · Topic

quantification

S. Lee · H. Yoo
Department of IT Policy and Management, Graduate School, Soongsil University, Seoul, Korea
e-mail: [email protected]
H. Yoo
e-mail: [email protected]
S. Seo
Hanwha Systems, Seoul 05451, Korea
e-mail: [email protected]
G. Gim (B)
Department of Business Administration, Soongsil University, Seoul, Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 95

J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_8
96 S. Lee et al.

1 Introduction

The term “Fourth Industrial Revolution” refers to the next-generation industrial revo-
lution led by Artificial Intelligence, the Internet of Things, robot technology, drones,
self-driving cars and Virtual Reality [1]. The term was first used by Klaus Schwab, the
chairman of the Davos Forum held in Switzerland in June 2016. There are many tech-
nical components of the “fourth industrial revolution,” and among them, the areas
related to digital content are “Virtual Reality” and “Augmented Reality.” When a
new concept or technology is first introduced, the media or researchers produce
numerous positive or negative information related to it. However, the actual assess-
ment or measurement of that information is discussed qualitatively again only when
the issue is raised many years later. It is only after the impact on technological and
social changes is re-examined with statistics on individual industrial areas or social
changes that it is accepted into the everyday life of society. As such discussions on
technical and social issues are not currently being measured in an immediate and
quantitative format, specific strategic responses to technical and social discourse at
this point are not easy. Moreover, there is a risk of consumingly spreading responses
to social and technological changes. More scientific solutions to prevent this would
be to define what will follow through quantitative measurements of social and tech-
nical discourse. For this reason, a strategic framework for social and technological
discourse is urgently needed, and this study aims to observe how each topic has
changed over time by automatically extracting and quantifying key topics from news
articles before and after 2016, the first public discussion point of the Fourth Industrial
Revolution on the subject of ‘Virtual Reality’ and ‘Augmented Reality’.

2 Background

2.1 Business Strategy and Text Analysis

Recently, Business strategy is a business administration subject that many researchers

have studied for a long time. It can be studied as a research topic in the realm of
strategy management or business intelligence and used as a function of commer-
cial software [2]. In the area of strategic management, there have been many cases
in which business strategy frameworks have been studied using text analysis over
the past decade. White et al. analyzed trends in the study of Information Strategy
Management (ISM) from 2000 to 2013 using text analysis and bibliography [3].
Seol et al. used the patented information to analyze and identify new business areas
using textual analysis techniques [4]. In addition, Lau et al. used the text analysis
method to study the application of the hotel industry [5], Fan et al. analyzed the text
analysis software and considered tools to help make strategic decisions [2]. A recent
meaningful study was an analysis of energy industry areas using LDA, a technique
Study on Business Strategy Quantification … 97

of topical modeling. It used predefined and quantified keywords to analyze trends

and status of enterprises or industries [6].

2.2 Text Analysis and Topic Modeling

In Many previous studies have been conducted on technology trend analysis using text
analysis. Some studies have been done on text analysis itself, but some researchers,
not computer-related researchers, have used specific techniques in text analysis to
analyze topics, trends, or social likes or dislikes on specific topics. In that way, the
main focus was simply to observe or extract trends in major keywords based on the
frequency of the emergence of key keywords in each period [7] and to make visual
representations through word cloud [8]. In some cases, the sentimental analysis
based on expressions was done [9]. In Korea, text analysis techniques have been
used as a way of analyzing research trends, especially in the areas of education
[10], policy [11], and economy [12] as well as trends in technology [13]. Topic
modeling, one of these methods of text analysis, is a statistical model for discovering
abstract “topic” of a set of documents [14]. In other words, it is a technique that
uses a statistical algorithm to clusters and analyzes them around topics that are key
to an unstructured set of documents [14]. In Korea, research trends on AI (Artificial
Intelligence) technology [13], research trends related to record management studies
[15], and research trends on FinTech technology [16] were also analyzed using topical
modeling. Topic modeling focuses on grouping documents with similar themes into
each cluster in a large set of unknown documents, with related techniques such as
LSI (Latent Semantic Indexing) and LDA (Latent Direchlet Allocation) [17]. In this
study, LDA was applied as a topical modeling technique, which is a technique that
deduces hidden topics from a set of documents using the probability of keyword
appearance in the document. This technique is based on the premise that when a
document is created, it is written with a keyword containing the subject.

2.3 Word Embedding

The traditional text analysis methodology used the frequency of the emergence of
noun-oriented keywords extracted through pre-processing in the analysis. Exam-
ples of the tradition methodology include the Information Retrieval Model [18] that
utilizes Term Frequency-Inverse Document Frequency (TF-IDF) as a Feature Value,
and word cloud [8] that visually processes the frequency of keywords to provide an
intuitive understanding of the nature of a group of documents. These techniques
assumed keywords extracted through preprocessing as variables and conducted
various analyses, but due to the existence of too many keywords in the nature of
text analysis, there was a limit to treating them as variables. Many studies have been
conducted to overcome this limitation, and in recent years, linear algebra and neural
98 S. Lee et al.

network methods have begun to emerge which does not make the keyword itself as
variable [19]. For example, it is a technique that vectorizes individual keywords of
the corpus given according to dimension N specified by the researcher into keywords
with dimension N. Word2vec and FastText are well-known examples of such tech-
nique. These techniques are called word-embedding in the sense that in the process
of producing human-specified N-dimensional vectors, the information in the context
in which keywords exist is incorporated into the N-dimensional to be vectored.
The advantage of word embedding method over traditional frequency-based
methods is that the efficiency of the operation is increased by first enabling a small
level of vector operation. Secondly, as the result of vectorization, similar words
between keywords can be detected in vector space as a method of calculating the
similarity between keywords. In this study, Word2vec is used among these tech-
niques. The similarity between each keyword of the topics obtained through the
LDA and any keyword representing the 2-dimensional axis (which is subsequently
referred to as the ‘strategic axis’) specified by the researcher was measured. After,
it was applied to quantify the topics by marking the location of each topic in the
2-dimensional strategic axis space.

3 Research Methods

This study analyzes social trends related to Virtual Reality and Augmented Reality
among digital content. In detail, it analyzes before and after the year 2016, when
the ‘fourth industrial revolution’ began to receive attention as a popular theme. The
analyzed data for service trend analysis used news data from 1 January 2011 to 7
June 2020. In using news data for analysis, the analysis was divided into two news
data as of 2016, when the media’s interest in the “fourth industrial revolution” began
to be noted. The two data sets are news articles from 2011 to 2015 and news articles
from 2016 to June 10, 2020, to see how social interest before and after that shifted
as of 2016.
Topic modeling was used as a method of analysis to secure topics that span the
entire duration of news articles, and LDA (Latent Direchlet Allocation) was used as
a specific technique. In addition, the keywords of the topics derived as a result of
the LDA were reviewed and the names of each topic were confirmed through the
survey. The correlation between each topic was then analyzed through an Inter-topic
Distance Map (IDM), which visually shows the distance over the two-dimensional
coordinate space [20].
In previous studies, research trends or policy trends were analyzed through IDM
analysis through the scale of each topic or the proportion of keywords contained in
each topic. This study took a step further from the existing research and analyzed
where each topic moved on the strategic axis by quantifying it on the strategic
axis using the Word2vec technique and expressing it on the strategic axis in two
dimensions. Figure 1 illustrates the experimental design of this study.
Study on Business Strategy Quantification … 99

Fig. 1 Design of the experiment

3.1 Data Collection

To study the social trends of VR and AR, the study utilized news data from BigKinds,
which is serviced by the Korea Press Foundation [21]. BigKinds is the nation’s
largest public news archive with the latest news from 54 major local media outlets,
including newspapers and broadcasting. It is being updated daily and accumulating
60 million news stories for over 30 years. BigKinds is suitable for data analysis
because it provides the original text of the news and a list of tokens extracted through
form analysis for each news article. In this paper, “VR”, “AR”, “Virtual Reality”, and
“Augmented Reality” were set as search terms using the search function of BigKinds,
and the period was set from 2011 to June 7, 2020, so that the trend of the four years
before and after 2016 can be seen. In addition, given that the topics to be covered
in this paper are digital contents such as VR and AR, but the base is on the field of
technology, the press company was limited to Digital Times and Electronic Times
only. The categories of news articles were limited to economics, society, culture, and
IT science. As a result, 14,443 news articles were searched in 2011, 283 in 2012, 226
in 2013, 452 in 2014, 924 in 2015, 3017 in 2016, 2772 in 2018, 2908 in 2019, and
779 in 2020. The news data downloaded in the form of the Excel program consists
of various columns for each news, such as date, media, classification, title, article
text, keyword, and character extraction. In this paper, only the ‘keyword’ column
was analyzed in consideration of the purpose of analysis.

3.2 Selecting Topic Title

Topics produced through the LDA are given a topic index expressed in numbers from
0 to 13 (the number of Topics K = 14). In the LDA technique, there was a study to
100 S. Lee et al.

automatically assign titles to each topic [22], but if the titles of the topics are expressed
numerically, it is difficult for humans to intuitively understand or recognize them. To
compensate for these shortcomings, this study conducted a survey on experts in the
IT field to give each topic a title. The method of the survey was shown by listing 10
keywords containing 14 topics and respondents looking at a set of keywords were
told to choose the appropriate title. Views presented as candidates for the title of
the topic consisted of two to three title phrases, randomly combining keywords that
formed the topic. The reliability verification that follows the title of the topic was
verified using the Intraclass Correlation Coefficient [6, 23] and the reliability analysis
function of SPSS 25.

3.3 Strategy Axis and Strategy Space

What sets this study apart from previous studies is that it defined the concepts of
strategy axis and strategy space. The term ‘strategic axis’ in this study was defined
as ‘a linear set of similarity values between the strategic keyword and the topic
keyword that exists within the word vector space’. Strategic space is defined as a ‘two-
dimensional plane represented by two strategic axes in a reciprocal relationship’. The
key idea of this study is to measure the similarity between each of the more than one
keyword related to strategy and the words that make up the LDA topics, calculate
the mean, quantify and observe the change in the strategic space. For example,
a bundle of words related to contrasting yet language-definable concepts, such as
‘opportunity’ and ‘threat’ and ‘strong’ and ‘weakness’ in SWOT analysis [24], can
calculate similarities with other words within the word vector space [25]. At this time,
the collection of similarity values calculated with the keyword ‘opportunity’, ‘threat’,
‘strongness’, and ‘weakness’ and the topical keywords in word vector space can be
called strategic axes. For example, the two-dimensional coordinate space expressed
by ‘opportunity’ and ‘threat’ can be defined as strategic space. These strategic axes
inevitably differ in the composition of keywords depending on the environment that
defines the strategic axis. If the strategic axis of ‘opportunity’ is assumed, the keyword
that constitutes the strategic axis in the new product launch will be the word related to
the actual expectations of the product, and the keyword that constitutes the strategic
axis of the ‘threat’ will be the keyword related to regulation according to the market
environment.

3.4 Word Embedding and Quantifying Topics

Word2vec was used as a word-embedding technique in this study. Word2vec is a

method of learning and vectoring tokenized keywords in consideration of the order
of appearance, as discussed in the consideration of prior studies. It is characterized
by learning in the order in which the words appear, including information on the
Study on Business Strategy Quantification … 101

meaning of the sentences in which the words were included. Word2vec was used in
two parts of this study. The first use was to select keywords to be used on the strategic
axis, and the second was to quantify topics. The selection of keywords to be used
on the strategic axis applied Word2vec for the entire duration of the news article,
and then found words like the word ‘public’ and ‘private’ in word vector space and
selected words that were actually semantically similar. When quantifying the topic,
each of the top 10 keywords that make up each topic and each keyword that forms the
strategic axis were measured in word vector space and the average of its similarity
values was taken as the corresponding strategic axis of the topic. This is expressed
in formula as follows:

When Topic K = {w1 , w2 , . . . , wn }

Strategic Axis S = {s1 , s2 , . . . , sn }Sx
n
i, j=1 sim(wi, s j)
Sx =
i×j

It is expressed in such formula which allows each topic to be calculated as a

corresponding value to one of the strategic axes.
Meanwhile, for this purpose of quantifying the topic, the news article data was
divided into data from January 1, 2011 to December 31, 2015 and data from January
1, 2016 to June 7, 2020 respectively. Keywords on the strategic axis for each word
vector corresponding to a divided period also measured the similarity in each word
vector space with different periods. This is because they assumed that the news
articles for the topics for that period might have different expressions. To make word
vector, word vector was created by first making news data for the entire period and
then dividing news articles into two data as of 2016. Both vectors, which are word
vectors for the entire period and when they are divided, are set to be the same.

4 Experiment Result

4.1 Topic Analysis

The results of the topic modeling are shown in Table 1 to extract the central topics and
related key words in news articles related to “VR” and “AR”. Each topic was presented
in order of proportion of topics using LDA-based topical modeling algorithms. The
top 10 words by frequency, representing each topic, were extracted. A random title
was then given around the association of the top 10 words by topic, and a survey was
conducted on experts to verify the objectivity of the title. The survey of 22 experts
surveyed 14 topics in the form of viewing keywords contained in each topic and
selecting the most appropriate title. The Intraclass Correlation Coefficient (ICC) for
102 S. Lee et al.

Table 1 Result of topic modeling on news articles

Topic Topic name Keyword
1 Government policy Industry, government, ICT, sector,
revolution, convergence, policy, drive, SW,
innovation
2 Smartphone camera Smartphone, product, camera, Samsung
Electronics, launch, device, Google, use,
Apple, video
3 Startup Company, support, business, startup,
start-up, investment, representative, www,
center, global
4 4th industrial revolution Ai, robots, services, iot, enterprise,
platform, data, cloud, future, digital
5 Global investment Market, business, investment, dollar, China,
growth, sales, patents, displays, prospects
6 Smart information solution Smart, information, solutions, utilization,
application, deployment, management,
automotive, possible, real-time
7 Mobile advertising service Services, mobile, delivery, customer,
information, smartphone, advertising,
products, use, SNS
8 5G service Service, 5g, network, standard, world,
Olympics, speed, movement, commercial
9 Education program Education, reality, students, universities,
learning, programs, utilization, virtual,
teaching, science
10 Video digital platform Content, production, culture, media, video,
digital, platform, business, realistic
11 Experience exhibition Experience, exhibition, space, visitors,
preparation, exhibition, booth, world,
operation, participation
12 Mobile game market Game, mobile, market, launch,
representative, online, Chinese, character,
pc, user
13 Carrier broadcasting platform service Service, KT, SK Telecom, TV, provision,
broadcasting, video, LGU+, content, LG
14 Event information Event, hosting, hosting, theme, Seoul,
introduction, participation, forum, industry,
com

the survey results was relatively high at 0.75, confirming the reliability of the topic
title selection through the survey.
Some of the calculated topical keywords such as com and www. seem meaningless.
Presumably, keywords extracted from the website’s address were not specifically
excluded because they were not thought to have a significant impact on the topic
assignment of topics.
Study on Business Strategy Quantification … 103

4.2 Change of Topic in Strategic Space

Next, the paper will examine the change in location of topics by timing in strategic
space. To express the strategic axis, ‘public’ and ‘private’ were first set as repre-
sentative keywords for each strategic axis. Similar words were searched in the word
vector space where the representative keywords you set were intended for the entire
time period. Among them, keywords similar to representative keywords were visu-
ally selected, and each added three keywords to create a set of strategic keywords.
The set of each strategic keyword is as follows:
strategy_word_X=[‘public’, ‘public institution’,
‘public sector’]
strategy_word_Y=[‘private’, ‘private institution’,
‘private sector’]
Next, a set of strategic keywords was quantified by calculating the individual
keywords contained in each topic. The values of the topics calculated for word
vectors can be expressed in the strategic space, as shown in Fig. 2. However, to avoid
the complexity of expression on the chart, only 10 topics out of 14 topics, which
account for 80% of the total keywords, were expressed in the strategic space.
As shown in Fig. 2, each topic was quantified according to the similarities between
‘public’ and ‘private’ and expressed in the strategic space, for example, in the case of
‘government policy’ topics located at the top right, the similarities between ‘public’
and ‘private’. This means that in all news articles since 2011, ‘government policy’
has a high similarity with both ‘public’ and ‘private’. On the other hand, the “global
investment” topic has relatively high similarities with “private” and much less similar-
ities with “public.” The limitation in reviewing the full-time data is that the overall
representation is similar to the post-2016 aspect because it exceeds the previous
volume of news articles related to the fourth industry since 2016.

Fig. 2 Visualization of strategic space on the whole data

104 S. Lee et al.

Fig. 3 Changes in topic in strategic space regarding ‘public’ versus ‘private’ before and after 2016

Next, quantified topics were expressed in strategic space by computing them with
word vectors before and after 2016.
Figure 3 shows the location of topics before 2016 as ‘* (star)’ and ‘o’ for topics
after 2016. Overall, service and platform-related topics are relatively little different
from “private” than before 2016, but the similarity with “public” has decreased
significantly. In the case of the “fourth industrial revolution,” the similarity between
“private” and “public” radios has increased, which is estimated to be based on the
public perception of the “fourth industrial revolution.” The ‘government policy’ or
‘start-up’ topics have not changed much in the post-war period.

5 Conclusion

This study conducted data analysis focusing on news articles extracted by searching
under the keywords “Virtual Reality” and “Augmented Reality.” About 14,000 news
articles from 2011 were analyzed through topic modeling, and various topics related
to “Virtual Reality” and “Augmented Reality,” which are central materials of this
study, were widely discussed in news articles. In particular, it was confirmed that they
were interested in the material base of digital contents such as telecommunication
and video, and topics like utilization services. It was also confirmed that there was
a topic on global investment due to the characteristics of digital content. Topics, the
result of analysis using the LDA technique of topic modeling, were named through
a survey, and reliability analysis was performed to ensure the reliability of the name.
The named topics were easy ways to examine the change in location of topics in
strategic space, as they were intuitively understandable to people. In addition, the
collected news article data was word vectored using Word2vec, one of the word-
embedding methods, and used to calculate the similarity between each keyword and
topic for the strategic axis of ‘public’ and ‘private’ specified in this study. As a result,
Study on Business Strategy Quantification … 105

news article data showed that topics related to service platforms had become much
distant from those related to ‘public’ compared to news articles before the ‘fourth
industrial revolution’ was publicized. The topics on ‘global investment’ showed the
same. However, the ‘education program’ topic and the ‘start-up’ topic was seen
to form a more similar distance to the ‘public’ at least in detail. In particular, the
‘education program’ was seen to have a slightly higher similarity in terms of ‘private’.
As shown in the research, this study is meaningful in quantifying and expressing
the relationship between each topic, which is the result of the strategic axis and the
topic modeling, in a two-dimensional space. It is expected that these techniques will
serve as an aid to the determination of future strategic and policy-oriented decisions
by the entity or government.

Future Challenges and Discussion Topics

This study showed that topics derived from topic modeling can be expressed in
strategic space both nominally and visually. Nevertheless, the space on the strategic
axis did not identify what factors caused each topic to shift at a point in time. This is
a future research project that relates to explainable artificial intelligence (XAI), and
requires a lot of research in the future. It is also necessary to apply the methodology
of this study to many already proven business strategy frameworks. This is expected
to serve as an opportunity to make wider use of this research’s automated business
strategy or text analysis as an aid to strategy establishment. These following topics
should be discussed in future in connection with this study:
• Under what conditions and statistical methods will the homogeneity of the sample
be demonstrated when there is a difference in the number of articles in time?
• How can scientific approaches and automation be realized in selecting keywords
for the strategic axis?
• Can topics be named automatically?
• How do we verify the change in the location of topics in the strategic space?
• Can’t we embed information other than the information obtained through the
sentences in the news article in Word Vector?
• What should be considered when comparing entities such as enterprises or
countries, rather than strategic pillars based on the concept?

In addition, the XAI and detailed research topics related to this study are expected
to capture when and why future changes in corporate or government policies can be
detected and responded to automatically in advance.
106 S. Lee et al.

References

1. Schwab K (2017) The fourth industrial revolution. Currency

2. Fan W et al (2006) Tapping the power of text mining. Commun ACM 49(9):76–82
3. White GO et al (2016) Trends in international strategic management research from 2000 to
2013: text mining and bibliometric analyses. Manag Int Rev 56(1):35–65
4. Seol H, Lee S, Kim C (2011) Identifying new business areas using patent information: a DEA
and text mining approach. Expert Syst Appl 38(4):2933–41
5. Lau K-N, Lee K-H, Ho Y (2005) Text mining for the hotel industry. Cornell Hotel Restaur
Adm Q 46:344–362
6. Pröllochs N, Feuerriegel S (2020) Business analytics for strategic management: identifying
and assessing corporate challenges via topic modeling. Inf Manag 57(1):103070
7. Hulth A et al (2001) Automatic keyword extraction using domain knowledge. In: International
conference on intelligent text processing and computational linguistics. Springer, pp 472–482
8. Cui W et al (2010) Context preserving dynamic word cloud visualization. In: 2010 IEEE Pacific
visualization symposium (PacificVis). IEEE, pp 121–128
9. Bakshi RK et al (2016) Opinion mining and sentiment analysis. In: 2016 3rd international
conference on computing for sustainable global development (INDIACom). IEEE, pp 452–455
10. Kim YH et al (2019) Research subject trend analysis on educational innovation with network
text analysis. J Educ Innov Res 29(1):91–116
11. Park JS, Hong SG (2016) Research of trends in Busan regional innovation policies using text
mining. Korean J Local Gov Stud 20(1):1–20
12. Song H-J et al (2013) Trend analysis of Korean economy in the economic literature by
text mining techniques. In: Proceedings of the Korean Society for Information Management
conference. Korean Society for Information Management, pp 47–50
13. Chung MS, Lee JY (2018) Systemic analysis of research activities and trends related to artificial
intelligence (aI) technology based on latent Dirichlet allocation (Lda) model. J Korea Ind Inf
Syst Res 23(3):87–95
14. Steyvers M, Griffiths T (2007) Probabilistic topic models. In: Landauer TK, McNamara DS,
Dennis S, Kintsch W (eds) Handbook of latent semantic analysis. Erlbaum, NJ. Information
science in Korea using topic modeling. J Korean Soc Inf Manag 30(1):7–32
15. Park JH, Oh H-J (2017) Comparison of topic modeling methods for analyzing research trends of
archives management in Korea: focused on LDA and HDP. J Korean Lib Inf Sci Soc 48(4):235–
58
16. Kim TK, Choi HR, Lee HC (2016) A study on the research trends in Fintech using topic
modeling. J Korea Acad Ind Coop Soc 17(11):670–81
17. Wikipedia, kr. Topic modeling. https://ko.wikipedia.org/wiki/%ED%86%A0%ED%94%
BD_%EB%AA%A8%EB%8D%B8
18. Baeza-Yates R, Ribeiro-Neto B (1999) Modern information retrieval, vol 463. ACM Press,
New York
19. Mikolov T et al (2013) Efficient estimation of word representations in vector space. arXiv
preprint arXiv:1301.3781
20. Sievert C, Shirley K (2014) LDAvis: a method for visualizing and interpreting topics. In:
Proceedings of the workshop on interactive language learning, visualization, and interfaces,
pp 63–70
21. Korea Press Foundation, kr (2017) Big kinds. https://www.bigkinds.or.kr/
22. Lau JH et al (2011) Automatic labelling of topic models. In: Proceedings of the 49th annual
meeting of the Association for Computational Linguistics: Human Language Technologies, pp
1536–1545
Study on Business Strategy Quantification … 107

23. Bartko JJ (1966) The intraclass correlation coefficient as a measure of reliability. Psychol Rep
19(1):3–11
24. Hill T, Westbrook R (1997) SWOT analysis: it’s time for a product recall. Long Range Plan
30(1):46–52
25. Shahmirzadi O, Lugowski A, Younge K (2019) Text similarity in vector space models: a
comparative study. In: 2019 18th IEEE international conference on machine learning and
applications (ICMLA). IEEE, pp 659–666
3D Printing Signboard Production Using
3D Modeling Design

Jungkyu Moon and Deawoo Park

Abstract As Seoul has become one of the world’s top 10 cities, there are many
demands for beautification and design for signboards. For business owners, the orig-
inality and publicity of the signboard are very important. In order to reflect these
social demands, research using quaternary calculation technology for sign making is
needed. In this paper, we design a 3D modeling signboard with slicing for 3D printing
using a computer graphic. After the design is confirmed, modeling and printing for 3D
printing are performed, and signboards are produced using acrylic, epoxy, and LED.
LED signboards are divided into LED signboards for electric signboards and LED
signboards for character channels. They display logo characters, etc., and research
and manufacture signboards with high visibility and excellent advertising effects
even at night. The signboard studied in this paper has the advantages of long life
and low failure rate, and the disadvantage of high cost is to be solved by using 4th
industrial technology.

Keywords 3D modeling · 3D printing · Fusion 360 · LED · Signboard · Slicing

1 Introduction

Signboards in the city’s business district and commercial district are scattered,
reducing the aesthetics of the city. As Seoul has become one of the world’s top
10 cities, there are many demands for beautification and design for signboards. For
business owners who use signboards, the originality and publicity of signboards is
very important. For these social demands, research using 4th industrial technology
[1] for sign making is needed. Fourth industries include 3D printers, drones, IoT

J. Moon · D. Park (B)

Department of Convergence Engineering, Hoseo Graduate School of Venture, Seoul, Korea
e-mail: [email protected]
J. Moon
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 109
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_9
110 J. Moon and D. Park

(Internet of Things), smart farms, robots, information security, AI (Artificial Intelli-

gence) [2], and big data. I will deal with 3D printers. Signage design and production
using the latest industrialization technology is an industrial technology that can reflect
the creativity of the entrepreneur and make the city’s aesthetics beautiful.
In this paper, 3D modeling is designed using computer graphics. Therefore, the
design that reflects the needs of the business owner can be changed. After the design is
confirmed, modeling and printing for 3D printing are performed, and signboards are
produced using acrylic, epoxy, and LED (Light Emitting Diode). There are various
types such as printed flat signboard, LED signboard (Channel), neon signboard, post
signboard, non-illuminated character signboard (SCSI), image monthly signboard,
label signboard, window selection, and signboard. Among them, LED signboards
are divided into LED signboards used for electric signboards and LED signboards
used as letter channels. As they display logos and letters, the visibility is high even
at night, so the advertising effect is excellent. The advantage is that the service life
is long and the failure rate is low, but the disadvantage is that it is expensive. Fourth
industrial technology is needed to further enhance the advantages and complement
the disadvantages.

2 Related Studies

2.1 3D Modeling

Modeling is Model + ing, which means modeling, and literally means making a
model when interpreted [3]. 3D Modeling refers to realizing 2D drawings in 3D.
As software that implements 2D drawing, software such as Adobe’s Illustrator
and Photoshop can be representatively. 2D Drawings created through Illustrator,
Photoshop, etc. can be saved as DXF (Drawing Exchange Format) files and used.
In order to print with a 3D printer, 3D modeling of this DXF file is required.
3D Modeling Program Software is required for this task. Typical examples are
Autodesk’s Fusion360, Inventor, Auto CAD, 3D Max, and Software such as Maya.
In this paper, we will cover Fusion360. Fusion360 is a software that integrates
CAD, CAM and CAE and is a Software capable of 3D Design & Modeling, Elec-
tronics, simulation, Generative Design, Documentation Collaboration, and Manu-
facturing [4]. Load the 2D Drawing worked on the Illustrator into the Fusion360
Software Work Space and proceed with 3D Modeling using 3D tools such as Extrude,
Revolve, Sweep, Loft, etc. When modeling, modeling should be carried out in consid-
eration of various factors such as material usage, printing time, and possibility of
shrinkage during printing [5].
3D Printing Signboard Production Using 3D Modeling Design 111

2.2 Slicing

Slicing means converting the 3D modeling file to G code so that the 3D printer can
recognize and print the 3D modeling data. G code is a programming language used
in numerical control, and is mainly used for computer aided manufacturing through
automatic control machine tools.
In a fundamental aspect, G-code is a language that humans use to command how
to make something on a computerized machine tool, and how to make it is defined
as where the machine tool’s tool moves and at what speed [6]. In a normal Slice
Program, you can set the quality of production, amount of material, and printing time
by setting layer height value, wall thickness value, infill density, printing temperature,
flow, speed, travel, cooling, support, build plate adhesion type, etc. And Mold, Tree
Support, Fuzzy skin, etc. Special functions can also be used.
The Slice Program includes 3D Slice Program software provided by each 3D
Printer manufacturer.
In this paper, we will use Ultimaker Cura, a 3D printing slice software that is
widely used worldwide. This program has recommended profiles obtained as a result
of thousands of hours of testing, and about 400 settings can be set through ‘Custom
Mode’, and the features and printing experience are improved through regular updates
[7]. Cura basically has a Viewport that visualizes 3D Modeling Files of STL, OBJ,
etc. You can adjust various options such as Layer Height, Line Width, Wall Thick-
ness, Infill Density, and Printing Temperature. You can change it to Korean in the
environment setting, so you can change it according to your convenience.
Clicking the Slice Button at the bottom right converts the 3D Modeling File in STL
format into a G-code File that commands 3D Printer to print. In the G-code File, the
temperature of the 3D Printer Nozzle is what, at what speed, and how much filament
is extruded as show in Fig. 1. We can look at the expected movement path and stack
structure through Preview, and depending on the production you want to print, you
can set brim, raft, skirt, etc. on the bottom surface. Through this, the bottom surface
of the production is not well adhered to the floor and is bent and floats, or the nozzle
moves and touches the printing production. The case of losing can be compensated.
In addition, when you press the Slice Button, the time for the production to be
printed and the required amount of filament are displayed at the bottom right, so you
can check the expected filament consumption and time as shown in Fig. 1. G-code
files can be printed by directly connecting to a 3D printer, or files can be moved by
putting them on a USB or SD card.

2.3 3D Printing

3D Printer [8] is a manufacturing technology that creates a three-dimensional object

by sprinkling a continuous layer of material, also called additive manufacturing [9].
For 3D Printing, 3D Modeling File, 3D Printer, and 3D Printer Materials in OBJ and
112 J. Moon and D. Park

Fig. 1 Slicing in Ultimaker Cura

STL format are required, and a Slicing Software Program to send 3D Modeling File
Data to 3D Printer is required. To do 3D printing, you first need a 3D printer. 3D
Printer is a device that makes a three-dimensional structure by identifying products
using CAD (Computer Aided Design) and printing materials layer by layer based on
the data [10].
This is how 3D Printer prints production.
• FFF (Fused Filament Fabrication)
• FDM (Fused Deposition Modeling, Method)
• DLP (Digital Light Processing)
• SLA (Stereo Lithography Apparatus)
• SLS (Selective Laser Sintering)
• PBP (Powder Bed and inkjet head 3d Printing)
• MJM (Multi Jet Modeling)
• LOM (Laminated Object Manufacturing)
• Etc.
It varies according to the printing material and method, but in this paper, we will
deal with 3D printer of FFF (FDM) method [11].
These are the materials used in 3D printers of the FFF (FDM) method.
• PLA (Poly Lactic Acid)
• ABS (Acrylonitrile Butadiene Styrene)
• NYLON, PVA, WOOD etc.
In this paper, we will deal with PLA (Poly Lactic Acid) [12].
3D Printing Signboard Production Using 3D Modeling Design 113

Fig. 2 CAFÉ ADAGIO in PARIS BAGUETTE, ABC-MART DXF file imported into Fusion360

3 3D Printing Signboard Production 2D Design and 3D

Modeling

3.1 3D Printing Signboard Production 2D Design

In 3D Printing Signboard Production, logos, characters, and fonts that individuals or

companies want can be implemented relatively easily in shapes that are impossible in
the existing industry, but designs that are difficult to apply due to complex processes
even if possible.
In this paper, we intend to show the process of 3D Modeling, Slicing, and 3D
Printing from 2D Design to PARIS BAGUETTE, ABC-MART, and MEDIHEAL,
which are large franchises that are actually commercialized by using 3D Printer. 2D
Drawing is necessary before 3D Modeling. People who need 3D Printing Signboard
Production usually have their own logo or characters representing the company, and
have their own logo and 2D DXF file of the character. Let’s import this into the 3D
Modeling Program Fusion360 Workspace as show in Fig. 2.
In the process of 3D modeling after importing a DXF file into Fusion360, an error
may occasionally occur. After executing the Fusion360 Sketch mode on this DXF
file, use the functions of Line, Spline, Project, etc. Do it as shown in Fig. 3.

3.2 3D Printing Signboard Production 3D Modeling

We are trying to 3D Model a newly sketched Drawing in Fusion360. For 3D

Modeling, there are various functions such as Extrude, Revolve, Sweep, Loft, Rib,
Web, etc. In order to use these functions, a profile is required. In Sketch, Line and
Line or Line and Spline etc. Line and line meet to form a surface, and proceed with
114 J. Moon and D. Park

Fig. 3 A new sketch on the CAFÉ ADAGIO in PARIS BAGUETTE, ABC-MART dxf file imported
into Fusion360

Fig. 4 Extrude drawing while holding a profile of CAFÉ ADAGIO in PARIS BAGUETTE

this surface. The Extrude function is to create a depth with the axis of Z in 2D drawing
consisting of XY and 3D modeling as if it was a protruding shape as shown in Fig. 4.
Extrude can adjust the angle value in addition to the function of grabbing the
profile and increasing it to the desired height as shown in Fig. 5.
The tools used and their values differ depending on the production method of 3D
Printing Signboard Production desired by the company. For example, when you want
to produce Acrylic 3D Printing Signboard, you need to proceed with 3D Modeling
considering the part to apply Acrylic and where to put the LED, and if you want
Epoxy 3D Printing Signboard Production, you need to understand the space and
principle to pour Epoxy well as shown in Fig. 6.
Or, if there is a 3D Modeling File, you can load it as it is and proceed with the
work as shown in Fig. 7. However, this may or may not be possible depending on the
nature of the 3D Modeling Program, so you should use the 3D Modeling Program
you need every time.
3D Printing Signboard Production Using 3D Modeling Design 115

Fig. 5 Adjusting the angle value of the extrude function

Fig. 6 3D modeling of CAFÉ ADAGIO in PARIS BAGUETTE, ABC-MART drawing

Fig. 7 MEDIHEAL 3D
modeling imported 3D
modeling file into Fusion360
116 J. Moon and D. Park

3.3 Slicing for 3D Printing Signboard Production Printing

When printing 3D Printing Signboard Production, the completeness depends on the

Slicing setting value. In addition, it is important to reduce the quality and printing
time in the printing of production. In general, if the printing speed is increased to
reduce the printing time, the quality will be lowered.
To have good quality while reducing printing time, you can use various functions
of Slicing. 3D Printing Signboard Production goes through a post-processing process,
but let’s unify the Layer Height to 0.2 mm. Sometimes the layer height is increased
to reduce the printing time, but the quality of production is very poor and there is
a high possibility of loss. Let the wall thickness be 1.2 mm. If the wall thickness is
thin, the quality may be degraded because the infill line is reflected outside. Reduce
the Infill Value to less than 10%. One of the advantages of 3D Printing Signboard
Production is that it looks hard but light in weight. By reducing the infill value to
less than 10%, the advantages of 3D printing can be utilized and printing time can be
reduced. Since Printing Temperature differs according to the type of filament used,
set the appropriate temperature. In general, I think you can increase the print speed to
reduce the printing time. That is theoretically correct. However, increasing the print
speed greatly reduces the quality of the production and increases the possibility of
loss at the same time. Therefore, the print speed is set at 50 mm/s. Another important
thing is Travel. When the nozzle travels, the filament can be retracted in the section
where the filament should not come out, which prevents the formation of spider webs.
Retraction can be set such as retraction distance, retraction speed, etc. This depends
on the 3D printer and 3D printing situation.
The problem here is the combing mode. When executing combing mode, the
nozzle moves over the printed area, avoiding the section that needs retraction. Then,
the moving time becomes longer and the printing time becomes longer, but the
quality of production can be improved by making less spider webs. Let the fan speed
of Cooling be 100%. If cooling is not performed, the surface may flow down or the
nozzle may be clogged, causing a risk of malfunction. Support depends on the case.
If production can be printed without support, you can proceed without setting. Brim
is recommended for the build plate adhesion type. This is a useful function to help
settle the production. After completing the settings, press the Slice button to proceed
with Slicing.

4 3D Printing Signboard Production Printing

4.1 3D Printing Signboard Production Printing

3D Modeling of 2D Drawing and sending Sliced G-code File to 3D Printer enables

printing as shown in Fig. 8.
3D Printing Signboard Production Using 3D Modeling Design 117

Fig. 8 MEDIHEAL being printed

4.2 3D Printing Signboard Production Post-processing

Method

The production printed with the FDM type 3D printer is made of a laminated structure,
so the laminated side is visible and the quality of the production is degraded. To
supplement this, a post-processing method is required. In that way, various techniques
such as a method using friction, a method using a chemical reaction, a method using
putty and poly, etc. are used. As a post-processing method of 3D Printing Signboard
Production, all technologies other than fumigation technology that utilize chemical
reactions that are not suitable for filament of PLA components are used as shown in
Fig. 9.
First, polish the laminated structure visible on the surface with relatively rough
sandpaper (around 200 rooms), then apply a Surfacer to cover the surface, and polish
again with medium rough sandpaper (around 400 rooms). It increases the roughness

Fig. 9 3D printing
signboard production using
sandpaper
118 J. Moon and D. Park

Fig. 10 MEDIHEAL post-processing with putty method

of the sandpaper, and if this operation is repeated, the laminated structure becomes
invisible.
If the size of 3D Printing Signboard Production required by companies is large,
you can work with putty as shown in Fig. 10.

4.3 3D Printing Signboard Production, LED, Acrylic

When the post-processing method process of 3D Printing Signboard Production is

finished, you can attach LED and Acrylic as shown in Fig. 11. The farther the distance
between LED and Acrylic is, the more beautiful and spread out, but the height of 3D

Fig. 11 3D printing signboard production led, acrylic, epoxy applied

3D Printing Signboard Production Using 3D Modeling Design 119

Printing Signboard Production increases, which is not good for aesthetics. Therefore,
it is necessary to make the aesthetic as beautiful as possible by controlling the optimal
height and light spreading effect.

4.4 3D Printing Signboard Production, LED, Acrylic

See Fig. 12.

Fig. 12 Constructed 3D printing signboard production

4.5 Comparative Analysis Table

The frame of the existing signboard is made of iron material, and the 3D Printing
Signboard Production is made of PLA material, so the existing signboard is rela-
tively heavy. Therefore, there are risks and labor costs are high in construction. The
production unit price includes various expenses such as material cost, labor cost,
utility bills, etc., but the frame of the existing signboard requires large equipment
and a large space is required accordingly. In addition, labor costs are high because
people work with the equipment themselves, and the work environment is noisy
120 J. Moon and D. Park

Table 1 Comparative analysis table between existing signage and 3D printing signboard production
Existing signage 100

3D Printing Signboard
80
Production 60
40
20
0
Weight Labor Originalit
Risk (%) price (%)
(%) costs (%) y (%)

Existing signage 100 100 100 100 30

3D Printing Signboard 30 40 40 50 100
Production

and messy. However, 3D Printing Signboard Production is produced by 3D Printer

directly by pressing the Print button, and there is no need to watch it. In addition, it
does not require a large space and the work environment is relatively quiet and clean.
Existing signboards can also be produced to give various shapes and effects.
However, the resulting opportunity cost is large and there are forms that are impos-
sible. However, in 3D Printing Signboard Production, if you consider the shape and
effect you want and perform 3D Modeling, then the 3D Printer only needs to print
as shown in Table 1.

5 Conclusion

Convergence with 3D Printer technology, one of the fourth industries, is very impor-
tant to complement and highlight the weaknesses and advantages of the existing
signboard production market. Through this, design that was previously impossible or
difficult has become possible, and the weight is relatively reduced, thereby increasing
safety and reducing labor costs. In addition, it is possible to achieve challenges that
were difficult to try because it can significantly lower the cost of materials, production
and construction. Through continuous research and development of various materials
and designs, the 3D Printing Signboard Production industry in which 3D Printing
technology is fused is expected to develop further. Furthermore, we will study 3D
Printing Signboard Production that combines AI and IoT.
3D Printing Signboard Production Using 3D Modeling Design 121

References

1. Younjung H (2017) A study on the artistic role of design according to the fourth industrial
revolution. Korean Soc Sci Art 31:445–455
2. Suyoun C, Suyeon H, Deawoo P (2020) Platform design and source coding of AI responsive
AR business cards. J Korea Inst Inf Commun Eng 24(4):489–493
3. Modeling. Namuwiki. https://namu.wiki/w/%EB%AA%A8%EB%8D%B8%EB%A7%81
4. Fusion 360. Autodesk. https://www.autodesk.com/products/fusion-360/overview
5. Sichang L, Gwangcheol P (2017) The ABS material characteristic compensation method for
reflection 3D printing modeling compensation method for the design prototype production. J
Integr Des Res 16(2):9–20
6. Jaewon L, Seokwoo S, Dosung C (2017) A study on slice program GUI design for FDM 3D
print. J Integr Des Res 16(1):21–30
7. Cura. Ultimaker. https://ultimaker.com/ko/software/ultimaker-cura
8. Yonggoo K, Hyunkyu K, Geunsik S (2019) Strength variation with inter-layer fill factor of
FDM 3D printer. J Korean Soc Manuf Process Eng 18(3):66–73
9. 3D printing. Wikipedia. https://ko.wikipedia.org/wiki/3%EC%B0%A8%EC%9B%90_%
EC%9D%B8%EC%87%84
10. Youngmoon S (2015) Types and applications of 3D printers. Polym Sci Technol 26(5):404–409
11. Sukhan J, Juil Y (2017) Study of thermo plastic elastomer filament printing using 3D printing
of FDM technology. Korean Soc Mech Eng 3299–3300
12. Jiae R, Sarang C, Jisoo P, Jihyo A, Jungmyoung L (2020) Changes in properties of 3D printing
filaments by extruding at different temperatures and lignin contents. J Korea Tech Assoc Pulp
Paper Ind 52(3):120–128
AI-Based 3D Food Printing Using
Standard Composite Materials

Hyunju Yoo and Daewoo Park

Abstract 3D printing is one of the ways to advance the technology of the 4th indus-
trial revolution. Instead of making a casting tool for the desired product, it directly
produces the product through 3D printing. 3D printing can produce customized prod-
ucts for each individual, so it is possible to construct a small smart factory. In partic-
ular, AI (Artificial Intelligence) technology learns and judges legal judgments, cancer
diagnosis, appropriateness judgments and standards for food ingredients, etc. that
humans used to derive results. In the era of COVID-19, 3D food printing becomes
an important turning point for non-face-to-face business and personalized business.
3D food printing is a technology that enables direct production of small quanti-
ties using 3D digital design and personalized nutrition data. However, the current
development stage of 3D food printing technology is only at the level of making a
product with a simple form or only one material, and separate material processing
is required to reach an appropriate level of print quality due to the printing charac-
teristics of various food groups [1]. In addition, there are not enough structured data
available for learning, and no prior development and indicators have been developed
for standard composite materials that can be applied to various foods to reach print-
ability. In this paper, we use AI machine learning to obtain adequate print quality
in 3D food printing. We study supervised learning, unsupervised learning, and rein-
forcement learning of AI machine learning, and design algorithms. In AI machine
learning unsupervised learning, conformity and non-conformity are determined, and
the result of the derived standard composite material value is applied to papers to
evaluate printing adequacy. Through AI machine learning reinforcement learning,
print aptitude is evaluated through rheological analysis, and big data values of various
food groups applied with standard composite materials are secured.

Keywords Artificial intelligence · 3D food printing · Printability · Standard

composite materials · Rheology · Hydrocolloids

H. Yoo · D. Park (B)

Department of Convergence Engineering, Hoseo Graduate School of Venture, Seoul, Korea
e-mail: [email protected]
H. Yoo
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 123
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_10
124 H. Yoo and D. Park

1 Introduction

Recently, the food industry is actively investing in biotechnology fields, such as

applying ICT convergence technologies such as animal tissue culture and protein
extraction, and food development applied from the mid-2010s. The market is
projected to grow by 9.5% per year to $17.8 billion by 2025 [2]. The future
key technologies of food manufacturing and processing can be summarized as
nanotechnology, biotechnology, and 3D food printing.
However, the current development stage of 3D food printing technology is only at
the level of making a product with a simple form or only one material, and separate
material processing is required to reach an appropriate level of print quality due to the
printing characteristics of various food groups [3]. In addition, there is not enough
structured data that can be used for learning, and there are no prior developments and
indicators for standard composite materials that can be applied to various foods to
reach printability, so no machine learning algorithm has been developed accordingly.
In order to overcome these problems, the purpose of this study is to design two
types of standard composite materials that can be applied to various food groups
in order to obtain appropriate print quality in 3D printing, and classify them into
10 steps according to the concentration value, and a machine learning algorithm.
Conformity and non-conformity are determined through the process, and the result
of the derived standard composite material value is applied to papers to evaluate the
print adequacy through unsupervised learning.

2 Related Studies

Until now, it has been pointed out that additive manufacturing processing technology
in 3D food printing manufacturing is difficult to apply due to the fact that the food
is composed of mixtures (carbohydrates, fats and proteins) and the inherent physic-
ochemical characteristics of these ingredients [4]. The printability of a food matrix
mainly depends on the rheological properties that are largely influenced by printing
parameters [5]. Considering the consistency, viscosity, and coagulation properties
of food ingredients, some food ingredients are stable enough to retain their shape
after extrusion and lamination, such as chocolate, sugar, pasta, cheese, and mashed
potatoes [6]. However, some foods such as rice, meat, fruits and vegetables cannot
be printed easily, and food hydrocolloids and transglutaminases have been applied
to many foods to improve extrusion and structural stability [7].
Soybean protein is a promising food ingredient in 3D food printing to improve
print quality, but the relationship between food protein and printability is very limited.
As a macromolecule essential for food structure, food hydrocolloids (polysaccha-
rides, proteins or lipids) can be regarded as the skeleton of food structures, and
have almost all processing, taste, nutrition and health benefits of food [8]. The
AI-Based 3D Food Printing Using Standard Composite Materials 125

future of food hydrocolloid research is worth looking forward to in the interac-

tion between hydrocolloids and other food ingredients, the design of future func-
tional food structures, and the regulation of interactions with the body [9]. 3D food
printing can be applied to various food industries as it can produce individual foods
with completely different tastes and flavors as well as ingredients of food. When
applied to various food groups, the development of standard composite materials
that can expect uniform print quality improvement is expected to be expanded to the
general food industry, such as the formation of a new food culture, not the level of
shaping food in three dimensions [10].

3 3D Food Printing System and Process

3.1 3D Food Printing System and Process

The 3D food printing system consists of a mechanical device, food material, and
a program for 3D printing implementation. The mechanical device can be config-
ured in various ways according to the shape and physical properties of the food
material discharged by dispensing for dispensing, but it must be designed and
manufactured through information on the viscosity and viscoelasticity that can
control the dispensing speed and quantity. The discharging system is divided into
melt discharging, liquid syringe discharging, and semi-solid extrusion discharging
methods, and the range of solid is in powder form, liquid has a viscosity of 5–100cPs,
and paste is suitable in the range of 500–40,000cPs.
In manufacturing food in 3D printers, selection of raw materials for food and
information on properties of raw materials are important factors in order to make
food raw materials printable. In addition, the pretreated food raw material must be
stably maintained after being laminated in a plasticization or melting state while being
supplied as a liquid or solid powder having flowable during the printing process. The
shape of food can be maintained through reversible processing, printing temperature
change, gelation and additives.
In this study, the 3D printing process was carried out using a self-developed
extrusion-based 3D printer shown in Fig. 1, and the prepared standard composite
material and food material are put into a syringe and moved to the nozzle tip to
continuously extrude, fusing the previous layer, and designed. As a dispensing device
for dispensing, a nozzle tip with a maximum volume of 60 ml and a diameter of
1.2 mm was used for 3D printing. All printing experiments were performed at room
temperature, and slicing was controlled with open source software of CURA 15.04.6
(Ultimaker BV, Netherlands).
126 H. Yoo and D. Park

Fig. 1 Extruded FDM Type

3D food printer model used
in the experiment

3.2 Materials

3.2.1 Types of Food Additives for Standard Composite Materials

Design

Unlike chocolate, which is extrudable and stable enough to extrude and maintain its
shape after lamination, such as chocolate, some foods such as sugar, pasta, cheese,
mashed potatoes and carbohydrates, fruits and vegetables cannot be printed easily,
and the use of food additives is essential to improve extrusion and structural stability.
The soybean protein rich in essential and non-essential amino acids used in this
study has excellent physicochemical and functional properties, and is a successfully
printed material to form porous scaffolds.
In addition, hydrocolloids play an important role in the structure, processing,
stability, flavor, nutrition, and health benefits of food, and thus are currently actively
studied in the field of food science and technology. In food materials of 3D food
printing, it is widely used for texture measurement by emulsification of liquid foods,
stability of dispersion, thickening, gelling, etc., and has a property that tends to have
a hydration layer by attracting water molecules around it because of its affinity with
water. Raw protein, starch, gelatin, agar, and beet pectin are among the representative
hydrocolloids.
In this study, standard composite materials include Soy Protein Isolate and hydro-
colloid compounds, and hydrocolloids are divided into two groups. Gelatin (gelatin
AI-Based 3D Food Printing Using Standard Composite Materials 127

or gelatine) is a type of protein that has a transparent color and is mainly added
to foods that give a chewy texture such as jelly because it has little taste and is
decomposed by proteolytic enzymes (proteases). Alginic acid is a polysaccharide
acid in the cell wall of brown algae. It is an acid contained in the cell walls of brown
algae plants such as seaweed, seaweed, and kelp. The refined product is in the form
of white powder. Alginic acid has a very wide range of uses. It is used as a paste
when dyeing fabrics, is used to increase viscosity in ice cream, jam, mayonnaise,
margarine, etc., and is also used in the production of lotions, creams, pills, and paper
[11]. Carrageenan is a rubber-like substance collected from red algae plants and is
used as a sticky material when making various foods such as chocolate, ice cream,
syrup, and cheese.

3.2.2 Standard Composite Material Design for Improving Printability

The composite standard material (hereinafter SCM) was prepared as shown in

Table 1, including Soy Protein Isolate and a hydrocolloid compound. It is divided into
two types: isolated soy protein and hydrocolloid (Type A: gelatin+alginic acid/Type
B: carrageenan+xanthan gum). The A-type sample is prepared so that sodium alginate
is first dissolved in distilled water for 2 h with a stirrer so that the final concentration
is 0.5%. Then, gelatin particles were added to the sodium alginate solution to reach
a concentration of 1.0/2.0/4.0/6.0/10.0%, and the mixture was incubated in a water
bath at 45 °C for 1 h. The following SPI powder was added to the sodium alginate
and gelatin solution so that the final concentration reached 2.0/4.0/6.0/8.0/10.0%,
and the sample names were SPI-GA1, GA2,…GA5. B-type is prepared so that the
final concentration is 0.5/1.0/1.5/2.0/2.5% by dissolving the xanthan gum solution
in distilled water with a stirrer for 2 h. Then, the carrageenan particles were added
to the xanthan gum solution to reach a concentration of 1.0/2.0/4.0/6.0/10.0%, and

Table 1 Standard composite material and distilled water mixing concentration design
A-Type Sample Name Alginic acid Gelatin SPI
SPI-GA1 0.5/99.5 1.0/99.0 2.0/98.0
SPI-GA2 0.5/99.5 2.0/98.0 4.0/96.0
SPI-GA3 0.5/99.5 4.0/96.0 6.0/94.0
SPI-GA4 0.5/99.5 6.0/94.0 8.0/92.0
SPI-GA5 0.5/99.5 10.0/90.0 10.0/90.0
B-Type Sample Name Xanthan Gum Carrageenan SPI
SPI-CX1 0.5/99.5 1.0/99.0 2.0/98.0
SPI-CX2 1.0/99.0 2.0/98.0 4.0/96.0
SPI-CX3 1.5/98.5 4.0/96.0 6.0/94.0
SPI-CX4 2.0/98.0 6.0/94.0 8.0/92.0
SPI-CX5 2.5/97.5 10/90.0 10.0/90.0
128 H. Yoo and D. Park

Fig. 2 Composition of standard composite material samples (left: SPI-GA, right: SPI-CX)

the mixture was incubated in a water bath at 45 °C. for 1 h. Finally, SPI powder was
added to the xanthan gum and carrageenan solution so that the final concentration
reached 2.0/4.0/6.0/8.0/10.0%, and each sample name was SPI-CX1, CX2,…CX5
(Fig. 2).

3.2.3 Preparing Dough

In order to perform 3D printing by applying the standard composite material to the

dough, the ingredient ratio was formulated as in Table 1 to determine the dough
formulation, and to investigate the effect of the standard composite material on the
printability of the dough. Were mixed at a mass fraction of 0.5, 1, 2, 3 based on the
dough.

3.3 Rheological Properties

The rheological properties of standard composite materials and dough doughs of

different compositions were analyzed using a rheometer (MCR 302, Anton Paar,
Austria) equipped with a sandblasting parallel plate (PP25/s) with a diameter of
25 mm and a spacing of 1.

3.4 Statistical Analysis

For statistical analysis, analysis of variance was performed using the SPSS software
package, and the significance of each sample was verified using the Duncan’s multiple
range test with p < 0.05 level.
Table 2 Dough formulas with standard composite materials (SCM)
Ingredients Dough with SCM (g/100 g)
0 2 4 6 8 10 2 4 6 8 10
0 SPI-GA1 SPI-GA2 SPI-GA3 SPI-GA4 SPI-GA5 SPI-CX1 SPI-CX2 SPI-CX3 SPI-CX4 SPI-CX5
Flour 45 43 41 39 37 35 43 41 39 37 35
Butter 20 20 20 20 20 20 20 20 20 20 20
Sugar 22 22 22 22 22 22 22 22 22 22 22
Milk 13 13 13 13 13 13 13 13 13 13 13
AI-Based 3D Food Printing Using Standard Composite Materials

Total 100 100 100 100 100 100 100 100 100 100 100
129
130 H. Yoo and D. Park

4 AI-Based Printability Analysis of 3D Food Printing

4.1 AI Machine Learning Analysis and Algorithm Design

When designing an algorithm for AI machine learning, the size, quality, characteris-
tics and available computation time of the data, and what you want to do with the data
are the main determinants. As shown in Fig. 3, supervised learning of AI machine
learning involves an input variable composed of previously classified training data
and a desired output variable, and by analyzing the training data using an algorithm,
a function that maps the input variable to the output variable can be found. In this
paper, the algorithm of the supervised learning algorithm of AI machine learning
uses a decision tree algorithm based on cases of mixing standard composite mate-
rials of 3D food printing, and the output value for the input value of 3D food printing.
Perform pattern extraction to determine.
In this paper, supervised learning is conducted by classifying 3D food printing
input value standard composite materials into 10 types, and the result is fitted by

Fig. 3 Design of AI algorithm for 3D food printing

AI-Based 3D Food Printing Using Standard Composite Materials 131

obtaining the difference between the predicted value and the actual value repre-
senting printing suitability and the square root of the mean through regression anal-
ysis. Divided by and nonconforming, values that do not indicate printability are
excluded. The standard composite material determined to be suitable is applied to a
dough prepared in advance to perform print aptitude evaluation, and the above print
aptitude degree value is applied to evaluate it. The standard composite material func-
tion maps new cases from the training data and judges the results. In addition, to apply
the unsupervised learning of AI machine learning, the 3D food printing aptitude is
improved through data clustering and density estimation. As a result, the degree of
3D food printing aptitude is classified into Low, Medium, High, and Super High, and
printing aptitude index is assigned. The unsupervised learning of AI machine learning
extracts D food printing output pattern values from multi-class classification. The
data of 3D food printing print aptitude is processed by standardization and normal-
ization process appropriate for the label for AI reinforcement learning, and the print
aptitude value is determined according to the change of type and density of standard
composite material mapped by Big Data pattern analysis. Generate evaluation infor-
mation (reward) that can be evaluated. It is possible to derive the printing aptitude
and standardized big data of 3D food printing through repeated learning through
regression by expanding the application of 3D food printing output information to
chocolates, fruits and vegetables, and protein.

4.2 Analysis of Print Suitability of 3D Food Printing of AI

Result Value

In order to find the optimum value of 3D food printing aptitude, AI supervised

learning, unsupervised learning, and reinforcement learning methods were used.
First, the measured values of shear modulus for the composite standard material
classified into 10 indexes were measured as 593.54–4328.29cPs as shown in Table 3.
A-type soybean protein, gelatin, and sodium alginate complexes exhibited higher
shear modulus values than those of B-type soybean protein, carrageenan, and xanthan
gum. The case that falls within the print suitability range of the viscoelastic modulus,
which is directly connected to the hazardous shape retention force, is 500–40,000cPs,
and the case that does not fall within the printability range is defined as Not Suitable.
Therefore, among the composite materials, SPI-CX1 and SPI-CX2 are excluded
from the print aptitude evaluation of standard composite materials. In addition, when
included in a suitable evaluation index, unsupervised learning is applied to Low for
10,000cPs or less, Medium for 10,000–20,000cPs, High for 20,000–30,000cPs, and
Super High for 30,000–40,000cPS. Such a process is standardized and applied even
if it is not a standard composite material designed to improve the print aptitude used
in this study, and is not suitable if it does not fall within the range of print aptitude
through supervised learning. Because it is marked as, big data can be built.
132 H. Yoo and D. Park

Table 3 Evaluation of printability of standard composite materials

Classification Sample name SCM concentration Shear modulus (Pa) Printability
(%) evaluation
A-Type SPI-GA1 2 593.54 Low
SPI-GA2 4 949.37 Low
SPI-GA3 6 1974.29 Medium
SPI-GA4 8 3354.97 Medium
SPI-GA5 10 4328.29 Medium
B-Type SPI-CX1 2 94.84 Not Suitable
SPI-CX2 4 397.59 Not Suitable
SPI-CX3 6 1002.54 Low
SPI-CX4 8 2557.39 Medium
SPI-CX5 10 3743.22 Medium

Table 4 Rheological properties and printability analysis of standard composite materials and dough
mixtures
Classification Sample name SCM concentration Shear modulus (Pa) Printability
(%) evaluation
A-Type SPI-GA1 2 3584.98 Medium
SPI-GA2 4 5854.33 Medium
SPI-GA3 6 12652.96 High
SPI-GA4 8 29772.33 High
SPI-GA5 10 41382.58 Super high
B-Type SPI-CX1 2 – –
SPI-CX2 4 – –
SPI-CX3 6 8492.99 Medium
SPI-CX4 8 19321.58 High
SPI-CX5 10 35326.22 High

Table 4 shows the result of measuring the shear modulus by applying the sample
classified as suitable in the print aptitude evaluation of the standard composite mate-
rial derived above to the dough. When various standard composite materials with
print aptitude values were applied, the standard composite material value was added
to the existing physical properties of the dough, resulting in a total elastic modulus of
3584.98–41,382.58cPs. When the same procedure as for the print aptitude evaluation
of the above standard composite material was performed, SPI-GA5 was excluded
because it deviated from the standard value of 41,382.58cPs. In addition, through the
evaluation of unsupervised learning, print aptitude values were classified from Low
to Super High.
AI-Based 3D Food Printing Using Standard Composite Materials 133

SPI-GA1 SPI-CX4 SPI-GA3 SPI-CX5

Low Medium High Super High
Fig. 4 Dough printout with complex standard materials

The printout is a dough printout showing the print aptitude status from Low to
Super High when the standard composite material is applied, and is a shape of a
cylinder having a diameter of 3 cm and a height of 6 cm, respectively (Fig. 4).

4.3 AI-Based 3D Food Printing Prospect

The optimal value of 3D food printing aptitude was calculated as A/B type 10 sample
values for AI supervised learning. The unsupervised learning of AI machine learning
is classified as a multi-class classification, and it is classified into Low, Medium, High,
and Super High to assign print aptitude indicators. 3D food printing reinforcement
learning method was used to extract appropriate values for rheological properties
and printability analysis of standard composite materials and dough mixtures.
The result based on the output of AI-based 3D Food Printing is shown in the figure.
When the basic 3D Food Printing method and AI machine learning proposed in the
paper are applied to supervised learning, unsupervised learning, and reinforcement
learning. The advantages are as follows (Table 5).

5 Conclusion

In this study, in order to improve 3D food printing printing aptitude, standard

composite materials composed of isolated soy protein and hydrocolloid materials
were classified into two types, and 10 samples were made with different concentra-
tion values, and applied to dough to improve printability. This work is to overcome
the limitation of 3D food printing, which cannot have standardized design values for
printability due to the variety of food raw materials and materials added to improve
printability. According to the existing 3D printing method, the print aptitude of the
sample applied to the standard composite material and dough was measured, and
according to the AI’s supervised-unsupervised-reinforced learning design model,
the appropriate-non-conforming, appropriate evaluation was performed. In future
134 H. Yoo and D. Park

research, it is necessary to apply various hierarchical factors that affect the satisfac-
tion of people’s food provided by 3D food printing and food ingredients through AI
deep learning reinforcement learning.

References

1. Kim C-T, Maeng J-S, Shin W-S, Shim I-C, Oh S-I, Jo Y-H, Kim J-H, Kim C-J (2016) Food
3D-printing technology and its application in the food industry. Korea Food Res Inst 2–10
2. Korea rural economic research institute: food technology status and challenges in the food
industry-focusing on alternative livestock and 3D food printing. Rural Econ Res Inst Basic Res
Rep 4–8 (2010)
3. Liu Y, Y Y, Liu C, Regensten JM, Liu X, Zhou P (2019) Rheological and mechanical behavior
of milk protein composite gel for extrusion-based 3D food printing. Food Sci Technol 102:338–
346
4. Derrossi A, Caporizzi R, Azzollini D, Severini C (2018) Application of 3D printing for
customized food. A case on the development of a fruit-based snack for children. J. Food

Table 5 Comparison of advantages and disadvantages of conventional 3D printing method and AI

application method
Comparison factor Division
Conventional 3D printing AI
method supervised-unsupervised-reinforcement
learning results applied
Printability evaluation Conclusions are drawn When learning values that can improve
through repeated various print aptitude such as food
experiments with a limited ingredient information, pretreatment
range of samples for status of food ingredients, additive
various food ingredients information, temperature, printing speed,
and additives etc. through the process of
guidance-unsupervised-enhancement,
optimal information values can be
obtained through vast amounts of big data
Application method of Various repeated It is possible to input information values
food additives experiments should be of various single and complex food
conducted to obtain limited additives that can improve printability,
information and quantified and obtain optimal printability index
indicators obtained values according to the additive
through a limited number information through the above process
of repeated experiments
for the type and amount of
additives designed by the
experimenter
Predicting the Material and physical When developing UX/UI by applying
implementation of the conditions must be set for and indexing data values for raw
outcome each individual situation in materials-food additives-physical
order to obtain optimal condition values maximization of user
results convenience
AI-Based 3D Food Printing Using Standard Composite Materials 135

Eng 65–75
5. Wang L, Zhang M, Bhandari B, Yang C (2018) Investigation on fish surimi gel as promising
food material for 3D printing. J Food Eng 220:101–108
6. Schutyser MAI, Houlder S, de Wit M, Buijsse CAP, Alting AC (2018) Fused deposition
modelling of sodium caseinate dispersions. J Food Eng 220:48–55
7. Guo Q, Ye A, Bellissimo N, Singh H, Rousseau D (2017) Modulating fat digestion through
food structure design. Elsevier 68:109–118
8. Lua W, Nishinarib K, Matsukawac S, Fanga Y (2020) The future trends of food hydrocolloids.
Elsevier 103:713–715
9. Ministry of science and technology information and communication: combination of advanced
3D printing technology and food tech! ‘3D Food Printer’ (2020)
10. Chen J, Mu T, Goffin D (2019) Application of soy protein isolate and hydrocolloids based
mixtures as promising food material in 3D food printing. J Food Eng 261:76–86
11. Fernandez C, Canet W, Alvarez D (2009) Quality of mashed potatoes: effect of adding blends
of kappa-carrageenan and Xanthan gum. Eur Food Res Technol 229(2):205–222
Telemedicine AI App for Prediction
of Pets Joint Diseases

Suyeon Han and Deawoo Park

Abstract Due to changes in lifestyles such as an increase in single-person house-

holds and the spread of Corona 19, the number of families with companion animals is
increasing. However, since companion animals cannot communicate with language,
measures for diseases or pain that cannot be seen with the naked eye are inevitably
insufficient. Among them, bone and joint disease affects the movement of companion
animals, and better treatment is possible if the joint disease can be detected in advance
in the stage before it worsens. In this paper, we intend to design a smartphone app
based on 5G communication that can diagnose the presence and possibility of bone
joint disease in companion animals. The Smarteck app takes photos of the back, side,
and front of the dog on a smartphone connected via 5G mobile communication and
uses the photographed photos and videos to have joint abnormalities such as patellar
dislocation, hip dislocation, and cruciate ligament rupture. It will be designed to
analyze and judge the presence of joint abnormalities through artificial intelligence
(AI) machine learning. The Smarteck app is designed to identify joint disease abnor-
malities and possible conditions and provide joint disease information and treatment
recommendations to dog guardian.

Keywords App · Artificial intelligence (AI) · Big data · Diagnosis · Smartphone ·

5G mobile communication

1 Introduction

With the development of the 4th industrial revolution technology, the use of smart-
phones through 5G communication is becoming popular. In addition, 5G commu-
nication smartphones are advancing based on AI that executes AR VR through

S. Han · D. Park (B)

Department of Convergence Engneering, Hoseo Graduate School of Venture, Seoul, Korea
e-mail: [email protected]
S. Han
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 137
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_11
138 S. Han and D. Park

high-speed video transmission and execution in real time and performs human
tasks such as real-time map route search, mobile e-commerce, and interpretation.
In particular, due to the COVID-19 PANDMIC incident in August 2020, there is
a national support and social demand for non-face-to-face telemedicine using 5G
communication smartphones.
As more and more people recognize companion animals as their families, interest
in treating companion animals’ diseases is also increasing. If non-face-to-face
telemedicine can be used to treat pet diseases, it will be possible to block COVID-19
infection pathways that can occur face-to-face, thereby preventing COVID-19.
Since companion animals and guardians cannot communicate in human language,
measures for diseases or pain that cannot be seen with the naked eye are inevitably
insufficient. Among the diseases of companion animals, joint disease is a disease that
affects the movement of companion animals. There may be a congenital joint abnor-
mality, or it may be a disease caused by an acquired cause. If the caregiver can detect
these joint diseases in advance, I think that the quality of life of the companion animal
will improve and the happiness index of the caregivers will increase by selecting
treatment at an appropriate time.
In this paper, we intend to design an AI smartphone app that can diagnose and
predict the presence and possibility of joint disease in companion animals. The
name of the app was named Smarteck by combining the English word “Smart” and
“Check”.
Smarteck App uses standards and data from professional veterinarians with more
than 20 years of clinical experience. We will build big data and store it in the cloud.
After running the Smarteck app on your smartphone, take photos or videos of
the back, side, and front of your dog, and match the photographed joint photos with
normal joint photos to determine the presence of joint abnormalities such as patella
dislocation, hip dislocation, and cruciate ligament rupture. do.
The Smarteck app is designed to determine the presence of joint abnormalities and
joint diseases and the probability of diseases of the dog, and to recommend related
treatments such as disease-related information, care method, and treatment timing
to the companion animal.

2 Related Studies

2.1 Machine Learning

Machine learning, one of the artificial intelligence technologies, came into the 2000s
and began to gain attention again. It is to change the parameters or structure of a
system so that computers can perform the same or similar tasks more efficiently
through experience and learning, and to create algorithms that use what they already
know to infer what they don’t know [1]. Machine learning methods can be broadly
divided into (1) supervised learning, (2) unsupervised learning, (3) reinforcement
Telemedicine AI App for Prediction of Pets Joint Diseases 139

learning, and (4) semi-supervised learning which is the middle between supervised
learning and unsupervised learning.
Supervised learning is a method of injecting data composed of pairs of problems
and answers into a computer to learn, so that when a similar new problem is given,
the answer can be found. Unsupervised learning is a method of training computers
to find answers by themselves and solve problems. Unsupervised learning includes
clustering to classify data with similar characteristics and learning association rules
to find relationships between characteristics in a large amount of data. Reinforcement
Learning is likened to the process of learning about the world when a human baby is
born. It is a method of finding a goal through the learning process, and a method of
determining a policy so that the reward is maximized through actions and rewards
accordingly in a specific environment.
In this paper, a dog joint abnormality diagnosis system app is designed using a data
mining pattern recognition method to predict joint diseases in companion animals.
Figure 1 shows the components and processing process of the pattern recognition
system. Pattern recognition systems are generally processed through five processes.
First, the object is measured in the real world, and the measured value is processed
to some extent to make it into a desired shape or normalized through a preprocessing
process that extracts a specific part.
After that, it is evaluated using an algorithm suitable for the type of problem, and
to verify that the evaluation is successful, it is verified through a model selection
process such as cross-validation or bootstrap, and results are derived [2].
Figure 1 shows the components and processing of the pattern recognition system.
Pattern recognition systems are generally processed through five processes.
First, the object is measured in the real world, and the measured value is processed
to a certain degree to make it into a desired shape or normalized through a pre-
processing process that extracts a specific part. After that, it is evaluated using
an algorithm suitable for the type of problem, and to verify that the evaluation is
successful, it is verified through a model selection process such as cross-validation
or bootstrap, and results are derived.

Fig. 1 Pattern recognition system components and processing

140 S. Han and D. Park

The Smarteck app in this paper is designed to be used in the steps before visiting
a veterinary hospital by analyzing dog joint photos and videos to determine joint
disease. It is necessary to build big data of dog joint photos and videos for analysis
and determination of the disease.

2.2 Dog Joint Disease

Patella dislocation: refers to a disease that comes out of or out of the patella lying
on the trochanter, and can be classified into 1st, 2nd, 3rd and 4th stages depending
on the degree and pattern of the dislocation. In the first stage, dislocation occurs
only when the patella is pushed inward or outward. Stage 2 can be intermittently
dislocated even when the patella is bent and stretched. However, it refers to a state
in which the patella returns to the normal position when the knee joint is moved to
the normal position or the knee joint is moved again. In stage 3, the patella is mostly
dislocated. Stage 4 means that the patella is permanently dislocated and cannot be
moved to its normal position. Most of the causes of patella dislocation are related
to musculoskeletal abnormalities such as medial quadriceps dislocation, the femur
twisting and bending outward, the femur end abnormally formed, the rotation of the
knee joint unstable, and the presence of deformities in the tibia [3].
Hip Dislocation: Hip dislocation is a disease that must be treated as soon as
possible to prevent continuous damage to the soft tissue surrounding the hip and
degeneration of the articular cartilage. When the hip joint is dislocated, it shows
claudication in which weight support is impossible. When hip dislocation occurs,
there is a difference in the length of the legs of both hind legs. The anterior dorsal
displacement is shorter in the affected leg than in the normal leg and is reversed in
the abdominal dislocation [4].
Cruciate ligament rupture: The cruciate ligament is the ligament that connects the
femur and tibia and serves to limit the anterior and posterior movement of these two
bones. It consists of an anterior cruciate ligament and a posterior cruciate ligament.
Cruciate ligament injury means that the ligament is partially or completely torn, or
the ligament that is attached to the bone is detached. If the anterior cruciate ligament
is partially ruptured, a mild pain response or lameness may be observed. Partial
rupture can proceed as a complete rupture over time [5].
Shoulder dislocation: It is not a common disease, but it refers to the separation of
the humerus and scapula due to loss or damage to a part of the structure supporting
the joint. It may be caused by trauma or may be congenital. Dislocation occurs when
the biceps tendon and the forelimb ligaments above the inner and outer joints are
torn or missing. He cannot support the weight of the dislocated leg and intermittently
walks with his leg bent [6].
Telemedicine AI App for Prediction of Pets Joint Diseases 141

3 Dog Joint Disease AI Machine Learning Design

3.1 Big Data Analysis and AI System Design by Dog Joint

Disease

In veterinary hospitals, treatment related to dog joint disease is generally conducted

in the order of “interview with guardian → gait observation → pain area promotion
→ radiography → radiation reading → diagnosis with veterinary opinion”.
In this paper, we use the pattern recognition method of data mining and design an
AI app for dog joint abnormality diagnosis system so that companion animal joint
disease can be predicted.
Table 1 shows the comparison between normal joint and medial and lateral patella
dislocation or hip dislocation disease suspicious shape for data construction.
The probabilities in the left column of Table 1 can be classified by the point used
when determining the presence or absence of a disease and the angle of the open
joint of the point, and it is designed to be built into cloud big data through supervised
learning of AI machine learning.
Points A and D denote the hip joint, B and E denote both knee joints, and C and F
denote both ankle joints. In a normal dog joint arrangement (green line) without joint
disease, the three joints above, hip joint, knee joint, and ankle joint lie on a nearly
straight line. However, the joint arrangement (red line) of a dog suffering from joint
disease goes out of line. The green line in Table 1 represents the normal arrangement,
and the red line represents the joint arrangement of the diseased dog.
The angle represented by the line in Table 1 is represented by an equation in
Table 2.

Table 1 Comparison of shapes for building big data by joint disease

Disease
Medial patella dislocation Lateral patella dislocation
Hip dislocation
(one side, two sides) (one side, two sides)
Percentage

50% ~ 90%
142 S. Han and D. Park

Table 2 Parameters for analyzing the presence or absence of joint disease (angle measurement
points)

f(s) = XL(θL1M, θL3M) + XL(θR1M, θR3M) + XL(θL1M, θL3M)+ XL(θR1M, θR3M)+ XL(θL1M,
θL3M) + XL(θR1M, θR3M)

θL1M = angle of red line AB and green line AB

θL3M = angle of red line BC and green line BC
θR1M = angle of red line DE and green line DE
θR3M = angle of red line DE and green line DE

3.2 Smarteck App Machine Learning Design

Design a smart app to be used in 5G communication smartphones. As a first step in

Smarteck App, we design an AI machine learning method using a rule-based expert
system. Figure 2 describes the five stages of a rule-based expert system. It can be
designed in 5 steps as shown in Fig. 2.
As shown in Fig. 2, a rule-based expert system is built based on expert knowledge
of veterinarians with more than 20 years of clinical practice. There may be some

Fig. 2 Rule-based expert system structure

Telemedicine AI App for Prediction of Pets Joint Diseases 143

differences depending on the dog species, but in general, dogs consider the joint
state in which the hip joint, knee joint, and ankle joint (ankle joint) are arranged
in a straight line as a normal model. Figure 3 shows the normal joint position and
arrangement. A and D represent the hip joint, B and E represent the knee joint, and
C and F represent the ankle joint.
First, these three joint points are determined as a reference, and the position and
arrangement are compared and analyzed to see how they differ from the normal
appearance. For example, in dogs with a patella dislocation, the knee joint (points
B, E) is pushed inward or outward, causing a deformation in the shape of the leg.
In dogs with hip dysplasia or hip dislocation, the arrangement of the hip, knee and
ankle joints deviated from a straight line when standing or walking due to a variation
of the hip joints (points A, D).
Establish the data so far as a data base, and standardize the data format and contents
by assigning a format to the data for each disease label. Based on the standardized
data format and knowledge, it makes rules to train computers. Table 3 shows how
the rule was applied to draw conclusions by implementing this rule as a rule to be
used when judging unilateral patella dislocation and creating a chain of reasoning.
The big data on the patella dislocation and hip joint dislocation constructed in 3-1
is used by a computer to learn big data through machine learning. DB and machine
learning compare and read the learned results, and access values of 50% or more
and 99% or less through error judgment (false detection) and excessive judgment
(overdetection).
The standardized data base is stored in the cloud through the Internet. The stored
cloud information is subjected to AI machine learning supervised learning to generate

Fig. 3 Dog normal joint

location and arrangement
144 S. Han and D. Park

Table 3 Rules to use for smarteck App

data for recognition and identification of joint diseases in dogs. Receives images of
dog joint disease from the smartphone smart phone using 5G communication. Joint
disease images are recognized by AI machine learning, identified, and transmitted
to smart phone smart app.
Figure 4 is a flow chart of Smarteck App. Run the Smartekc app on your smart-
phone. Take photos or videos based on the shape of the dog’s joints, and determine
the presence or absence of a disease through the Smarteck app. When it is deter-
mined that there is a disease by removing error judgment and excessive judgment
and selecting AI judgment, information and treatment related to disease and care are
recommended.

3.3 5G Communication Design on Smarteck App

Figure 5 shows the communication design of Smarteck App.

The higher the quality of videos and photos, which are the data necessary for
diagnosis, the more accurate it is to determine the presence or absence of a disease.
To this end, it utilizes large-capacity, high-speed transmission of 5G communication,
and improves the efficiency of image recognition and judgment on joint diseases with
the big data of the cloud and the learning result of AI machine learning.
Telemedicine AI App for Prediction of Pets Joint Diseases 145

Fig. 4 Smarteck flowchart

Fig. 5 Smarteck communication system

146 S. Han and D. Park

4 Design of Smarteck App for Determining the Presence

of Joint Disease in Dogs

Install the Smarteck app on a 5G communication smartphone, sign up as a member,

log in, and run the app. Figure 6 is the full screen configuration of the Smarteck app.
After logging into the smart phone smartphone app, photograph the dog and scan the

4.1 Shooting Mode

joint shape. AI machine learning compares with the supervised learning data, reads
whether the joint points match, and displays the read result on the screen.
As shown in Fig. 7, after running the Smarteck app on a 5G communication
smartphone, access the camera app and take a picture or video of the dog. When
taking pictures, take pictures so that both hind legs come out correctly. After shooting,
upload photos and videos to the Smarteck app.

4.2 Scan Mode

The captured two hind legs are displayed on the screen and displayed by touching
the app as shown in Fig. 8 based on the data of the hip joint, knee joint, and ankle
joint of AI machine learning that are the standard for comparative analysis.

Fig. 6 Smarteck App

Telemedicine AI App for Prediction of Pets Joint Diseases 147

Fig. 7 Shooting mode

Fig. 8 Scan mode

148 S. Han and D. Park

Fig. 9 Match/readout mode

4.3 AI Readout Mode

The information learned in AI machine learning is compared with the photographed

joints. Compare and analyze the normal joint model and the point taken by touching
in Fig. 8 as the pattern matching method.
Figure 9 is a screen under comparative analysis of dog joint disease.

4.4 Result Mode

The screen displays 4 items as the result of readout.

• What percentage does the reading match the shape of the normal joint?
• What percentage is the likelihood of having joint disease?
• What are your recommendations?
• Click for information on related joint diseases
Figure 10 is the result screen of the Smarteck app.
Telemedicine AI App for Prediction of Pets Joint Diseases 149

Fig. 10 Result mode

5 Conclusion

There is a social demand for non-face-to-face telemedicine with COVID-19

PANDMIC. In this paper, in order to solve the frustration of being unable to commu-
nicate about companion animal diseases through non-face-to-face telemedicine, we
designed an AI machine learning method for predicting dog joint disease through
the AI app Smarteck on a 5G communication smartphone.
Save the standardized data base in the cloud. Cloud information is based on AI
machine learning supervised learning, which creates the base data for recognition
and discrimination of joint diseases in dogs. Computers learn normalized and stan-
dardized big data for patella dislocation and hip dislocation of dogs using AI machine
learning. Through 5G communication, images of dog joint diseases are transmitted
from Smarteck App. Joint disease images are recognized by AI machine learning,
identified, and transmitted to smart phone smart app. The learned results are compared
and read by a computer using AI DB and machine learning supervised learning, and
error judgment and excessive judgment are eliminated by AI deep learning, and the
disease diagnosis probability is 50% or more and 99% value is approached.
Run the Smarteck app on your smartphone. A photo or video is taken based
on the dog’s joint shape, and the presence or absence of a disease is determined
through the Smarteck app. It was designed to remove error judgment and excessive
judgment, select AI judgment, and recommend disease and care-related information
and treatment when it is judged that there is a disease.
150 S. Han and D. Park

While designing the Smarteck app, it was found that research data on the compar-
ative angle of the normal joint position of a companion animal and the position of
a diseased joint were insufficient. It is necessary to study future comparative angles
by using AI deep learning technology for photos and videos that will be obtained
through the Smarteck app.

References

1. Kun-Myung L (2019) Artificial intelligence-from turing test to deep learning saengneung

publishing Co., Ltd. Gyeonggi-do
2. Bishop CM (2009) Pattern recognition and machine learning, Springer
3. Hamilton L, Farrell M, Mielke B, Solano M, Silva S, Calvo I (2020) The natural history of canine
occult grade II medial patellar luxation: an observational study. JSAP 61(Issue4):241–246
4. Chun-ki C, Seok-joong K, Ji-yeon L, Jae-hoon L (2020) Dog home training, pp 97–99, Crown
5. Chun-ki C, Seok-joong K, Ji-yeon L, Jae-hoon L (2020) Dog home training, pp 93–94, Crown
6. Chun-ki C, Seok-joong K, Ji-yeon L, Jae-hoon L (2020) Dog home training, pp 103-104, Crown
Design of Artificial Intelligence for Smart
Risk Pre-review System at the KC EMC

Youngjoo Oh and Deawoo Park

Abstract The world’s major countries use Free Trade Agreements and protec-
tionism to politicize non-tariff barriers along with tariffs that benefit their own
industries. When it comes to world trade, there are a total of 79 legally mandatory
certification systems in Korea. As part of the non-tariff barrier policy, electrical and
electronic products are subject to pre-evaluation of the “Broadcasting and Commu-
nication Equipment Conformity Assessment” system. In this paper, we design Arti-
ficial Intelligence, Big Data, and Cloud Computing platforms to improve existing
document evaluation. It analyzes big data in the form of electric and electronic prod-
ucts, and judges whether the pre-evaluation for the suitability of KC Electromagnetic
waves intended to be sold in Korea is valid through AI machine learning. In addition,
if it is not suitable, we intend to design an AI deep learning system that provides a
complementary measure and provide it in the form of a Cloud Computing platform.
We intend to develop a semi-automatic AI connection system that can perform KC
conformity assessment for electric and electronic products to be sold in Korea.

Keywords Artificial intelligence · Big data · Cloud computing platforms ·

Conformity assessment · Free trade agreements · Harmonized system code

1 Introduction

This paper proposes a design of an AI that reviews the conformity assessment of a

product in advance. In addition, the design model is explained by applying the laws
on “KC Broadcasting and Communication Equipment Conformity Assessment” of
Republic of Korea.

Y. Oh · D. Park (B)
Department of Convergence Engineering, Hoseo Graduate School of Venture, Seoul, Korea
e-mail: [email protected]
Y. Oh
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 151
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_12
152 Y. Oh and D. Park

Currently, there are 79 mandatory legal certification systems in 16 ministries

and 107 voluntary statutory certification systems in 24 ministries in Republic of
Korea. Even for the same purpose of “product safety” and “quality and performance
assessment”, there is an inconvenience of having to obtain duplicate certification
because each department has different certification marks. As a result, not only
time and money were wasted, but also international credibility was degraded, and
local wealth leakage caused problems such as the need to re-certify because mutual
authentication was not possible in cross-border transactions. Accordingly, 13 legally
mandatory certification marks are integrated into one national integrated certification
mark (2011.1). The initial national integrated certification mark (KC mark) was
launched by combining 10 certification marks from the Ministry of Trade, Industry
and Energy and the Ministry of Employment and Labor in September 2009. As of
November 2011, 13 certification marks from 5 ministries were integrated into the
KC mark. Has been. Since then, as each department established or changed the legal
obligation certification system, the KC mark was introduced, and as of September
2017, 23 legal certification systems in 8 departments are using the KC mark.
In particular, the laws applied in this paper are operated under the “Radio Waves
Act” and its subordinate laws as the “Broadcasting Communication Equipment
Conformity Assessment System (Conformity Certification, Conformity Registra-
tion, Provisional Certification)” under the jurisdiction of the Ministry of Science and
ICT, Technology and Communication. The products covered by this certification are
electrical and electronic technical equipment and components, and some machinery.
These products can be classified according to the Harmonized System Code. Perhaps,
most likely, it will be included in the items from Part 16 to Part 20. However, as the
market of IoT systems of the 4th industry expands, wireless communication functions
are fused to other items, which are applicable to this law. The convergence of these
products is making it difficult to determine the technical application and suitability
of this Act. As such a solution, a “Smart Risk Pre-Review System” is used to judge
suitability through preliminary review of the product. And check and supplement
problems in advance. This will help you enter the market. This rule of review also
applies to export products. As a result, it helps both domestic and foreign markets.
Because, the technical basis of the “KC Broadcasting and Communication Equip-
ment Conformity Evaluation System” follows the conformity evaluation system
of IEECC (IEC System for Conformity Assessment Schemes for Electrotechnical
Equipment and Components) under the International IEC (International Electrotech-
nical Commission). Therefore, it may be easier to learn this AI further and expand
its scope so that new products developed in Korea can be verified for the suitability
of the international market.
Since the launch of the Trump administration in 2017, the United States has
been around the world, including major allies. After the first 100 days, which
caused numerous controversies, the Trump administration has attempted to change
its foreign economic policy under the name of “America First!”. The ‘Pivot to
Asia’ and the ‘American Pacific Century Initiative’ were virtually ignored, and the
Trans-Pacific Strategic Economic Partnership(TPP) withdrew from the Free Trade
Agreement(FTA) for the first time in the history of US foreign policy without any
Design of Artificial Intelligence for Smart Risk … 153

congressional approval. Trump attacked immigrants from Mexico and promoted

the United States-Mexico-Canada Agreement (USMCA) through amendment of
the North American Free Trade Agreement. A ‘trade war’ was unfolded, involving
simultaneous battles with several countries, including traditional allies in Asia and
Asia.
The Trump administration abandoned its existing strategy to reconcile imbalances
within a multilateral framework and adopted a bilateral strategy that puts pressure on
individual states by focusing on the issue of the US balance of payments imbalance.
The U.S. attacked traditional allies such as the European Union based on the unusual
logic of placing a “national security threat” on trade issues, and a strategy of unilat-
erally imposing special tariffs on China without using the World Trade Organization
based on Article 301 of the Trade Act of 1976. Took. The “trade war” between the
United States and China lasted for more than two years, gradually having negative
effects on the global economy, as well as a profound political and economic risk
of the absence of a leading state in the world economy. In accordance with the US
President Donald Trump’s “America First” policy, the US government and China
are in a trade war. This trade war is an economic war between the United States
and China to preempt the economic advantage at the point of entering the stage of
economic integration. As a result of the post-economic war, the world is expected to
accelerate toward an integrated economic market. To take the initiative, a policy to
reinforce tariff and non-tariff trade barriers that impose tariffs equivalent to 10–30%
on each other is exchanged [1].
However, the trade war between the United States and China is causing a lot of
damage to each other and adversely affecting neighboring countries. For this reason,
this consuming trade war will soon end, and it is believed that economic integration
will occur rapidly, centering on the interests and interests of the country. In partic-
ular, the US, which has chosen a bilateral strategy from a multilateral strategy, will
achieve economic integration around the Free Trade Agreement, which represents
bilateralism.
Now, we live in a different world than before, where new technologies such as 5G
and cloud emerged in the era of the Fourth Industrial Revolution, where all people, not
a few, simply and easily find and share information. As a result, each country on the
planet has come to an era called the global information revolution. Powerful countries
with such technological advantages have entered into a Free Trade Agreement that
meets their own interests and interests, leading to protectionism. It is choosing a
policy to lower the tariff barrier that used to protect its own industry. The Free Trade
Agreement is the removal of various barriers to protection such as tariffs, which are
trade barriers for free trade between the two countries. This has the advantage of
allowing more free trade and exchange of products between countries, but there are
many in Korea in that countries with weak industrial structures are concerned about
collapse of the industrial sector and that countries with large amounts of capital and
technology dominate the culture of the other country. It is controversial. Countries
that have signed free trade agreements are building non-tariff barriers to protect their
industries to avoid this situation. However, these non-tariff barriers dominate most of
the market in large corporations that dominate capital and information and provide a
154 Y. Oh and D. Park

reason for preventing SMEs from entering the market. The “Smart Risk Preliminary
Review System” in this paper is a platform that provides and shares solutions for
the Technical Barrier Trade (TBT). It is expected that this platform will perform the
function of providing information democratization by sharing regulatory information
of products.
Figure 1 is a flow chart of the “Smart Risk Preliminary Review System” applying
the technical regulations and laws of the “KC Broadcasting and Communication
Equipment Conformity Assessment System”. It is designed to access and authenti-
cate information of subscribers in the production, manufacturing, distribution, and
customs fields of products using a block chain. Big data created in the “Smart
Risk Preliminary Review System” is designed to be shared and operated in a
cloud computing network. In the “Smart Risk Preliminary Review System”, the
HS code (Harmonized System Code; Harmonized Community Description and
Coding System) is designed to use AI machine learning, and the search for technical

Fig. 1 The flowchart for smart risk pre-review system in the KC EMC
Design of Artificial Intelligence for Smart Risk … 155

regulations and standards is designed to use AI deep learning algorithms. In addi-

tion, conformity verification AI machine learning can be learned, and conformance
completion DB is designed to apply safety/quality labeling rules.
“Smart Risk Preliminary Review System” is a non-face-to-face conformance
tester of products. This is because all AI engines are implemented with cloud
computing. The AI design for the smart risk preliminary review system studied
in this paper will be used as basic data for the development of public and private
institutions related to trade.

2 Analysis of Smart Risk Preview System

To sell products or to sell imported products, technical regulations, standards,

conformity assessment procedures, and safety labeling are defined in the importing
country’s legislation, as shown in Fig. 1. Must be checked for a conformity assess-
ment. These product sales conformity assessment processes are defined in interna-
tional standards, so that the Member State shall apply them in accordance with this
standard.
There are compulsory implementations of technical regulatory requirements. In
order to obtain the solution desired by the Smart Risk Pre-Review System, the
technical information of the product manual, the parts list, the circuit diagram,
the compounding cost, the manufacturing process diagram, the quality control
documents, and the cyber security of the use system are required. These product
preliminary reviews are digital documents written electronically that may be forged,
tampered with, or compromised. In addition, as product preliminary review informa-
tion is managed and distributed online, it is difficult to check the document history,
thereby ensuring authenticity, integrity, reliability, and usability of digital documents.
It must be necessary [2].

2.1 Digital Documents and Cybersecurity for Smart Risk

Preview

In order to obtain the solution that users want to the” Smart Risk Pre-review System”,
technical information and manufacturing processes of products, such as instruc-
tion manual, component list, circuit diagram, and mix ratio, and cybersecurity for
the system of users and security of quality control documents is required. This
product pre-review information are digital documents written electronically, poten-
tially tampering, and damaging. Furthermore, as product pre-review information is
managed and distributed online, it is necessary to obtain authenticity, integrity, reli-
ability, and usability for digital documents because document histories are difficult
to verify.
156 Y. Oh and D. Park

2.1.1 Blockchain

Blockchain was first published by an anonymous developer named Satoshi Nakamoto

in October 2008 in paper titled “Blockchain: A Peer-to-Peer Electronic Cash System
[3].”, in the cryptographic technological community “Gmane”. The paper explained
that double payments can be prevented by using P2P network link. Blockchain was
selected as one of the 10 key technologies of the 4rth Industrial Revolution at the
World Economic Forum in 2016. and they explained that the economic seize of the
Blockchain will account for 10 percent of GDP [4]. As shown in Fig. 2. Blockchain
consist of a header and body for each block. To prevent to falsify of block information,
it includes electronic signature technology of participants in the encryption network
of hash value data and in the encryption network key which is optimized for cyber-
attack defense. In the Blockchain, hash values are converted hash values that are
combined with the number of random specified lengths and characters to be used to
determine falsification [5]. Blockchain can be used in various fields such as financial,
economic, legal, electronic, voting public data management, etc. On the financial side,
it is expected create new financial instruments by combine pitch and blockchain
to P2P individual transaction method, not banking transaction method, activating
cryptocurrency money, cryptocurrency exchange, distribution, automatic investment,
etc.

Fig. 2 Blockchain structure

Design of Artificial Intelligence for Smart Risk … 157

As the central organization does not need function, the system will be newly
constructed based on reliability using a P2P based on blockchain, which will have
effect of maintenance cost and commission reduction in financial transactions. Also,
a new digital market will be created, such as e-voting and public data management.
With the development of Artificial Intelligence and Cloud computing, The 4rth
Industrial Revolution will bring about a major change in the social structure and
actively promote the application of blockchain technology in the development of
platform, device, application to prepare for the ripple effects of various industrial
institutions in a centralized structure, thus leading the future ecosystem by developing
new business creation and services, and is creating new revenue through contin-
uous improvement [6]. However, there is a disadvantage that blockchain applies and
storage cost increase as much as the number of users due to distributed storage of
data, and the cost and the time of network processing are needed to send a computing
power and a data to multiple locations simultaneously to store the same contents.

2.1.2 Blockchain Platform Classification

Blockchain which is currently being developed is built based on various platform for
each location and is typically divided into Private and Public types. This is developed
by utilizing the characteristics and scope of the blockchain participation network [7].
The Public Blockchain is a structure that anyone can allows participation in the
network of who is responsible for the blockchain without the permission of the
specific agency. The blockchain can be freely trade within the trade network. By using
the consensus algorism such as PoW (Proof-of-work) and PoS (Proof-of-Stake), all
users who participate in the public blockchain are shared all their trading books
to ensure high transparency [8]. These features can have perfect security even for
cyber-attacks, such as hacking, but slow network speeds, resulting in slow transaction
speeds.
The Private Blockchain is a centrally blockchain unlike the public blockchain
which does not share transaction details to all network users but provide high speed
processing by closed companies and participation blocks. If the central server is
under cyber-attack due to central control, all blockchain stored in server may not be
recoverable the person responsible for cyber-attack has unclear shortcomings. The
typical private blockchain is Hyperledger Fabric enveloped by Linux Foundation.
Hyperledger Fabric is composed by the share ledger, personal information, consensus
and smart contact. The composition of the share ledger is two share ledgers. This
consists of a Transaction Log that records all transactions and the world state that
stores a specific state [9].
Information is a comprehensive concept that includes information of various
electronic types, such as images and voice, in addition to information of docu-
ment types prepared by computer or information processing equipment. The over-
seas standard guidelines related to electronic document are ‘IOS 15489’ (Recorded
Management Procedure and Method), which are international standards for records
and management [10]. It was established under the supervision of ‘SC 11’ (Record
158 Y. Oh and D. Park

Management sub-committee) of ‘IOS/TC46’ (Information Technology committee).

The posed Chapter is to protect all records and standardize the management process
and procedure for more effective retrieval of evidence and information contained in
records. ‘IOS 15489’ classifies the quality requirements of records management as
“Progressive”, “Reliability”, “ Integrity”, “ Availability” [11].

2.2 Analyze the Process of Product Sales Conformity

Assessment

To sell a product or to sell an imported product, technical regulations, standards,

conformity assessment procedures and labeling as specified in the laws of the
importing country, such as Fig. 3. The product conformity assessment processes
are defined by International Standards, so that Contracting States shall apply them
in accordance with this standard in accordance with their own country.
The requirements for technical regulations are established by the Radio Wave Act,
the system and the government led by regulations that describe product characteris-
tics, related processes, and methods of production. The performance of the technical
requirements is mandatory. If the requirements are not complied with, the company
shall be fined, collect goods and be subject to legal restrictions.
Product characteristics, related processes and ISO standard production methods
follow the standardization conditions. For the common and repetitive use of stan-
dards, certification bodies such as KTL, UL and TUV give approval. There is no coer-
cion in the implementation of the standard. Producers and consumers gain reliability
from the market by getting the standard certification.
The process of ensuring that the product meets the technical requirements of
regulations and standards for the EMC is the conformity assessment. The conformity
assessment process includes testing, testing, and certification.
The quality indication is to print the product’s trademark, quality, content, etc. on
the package or to mark it by affixing it. The reliability of the product is notified to the

Fig. 3 The smart risk pre-review process

Design of Artificial Intelligence for Smart Risk … 159

consumer by disclosing information about the suitability of the product’s technical

regulations and standard requirements to the consumer. Quality displays are subject to
fines and mandatory warehousing because they set regulations in laws and technical
standards.

3 AI Design of Smart Risk Pre-review System

3.1 The Machine Learning Design to Search HS Code

The Harmonized System Code (Harmonized System Code; Harmonized Community

Description and Coding System) is a numerical code of the types of goods traded
in foreign trade transactions, which was established by the International Convention
in 1988. Basically, it consists of six digits, and each country is using four additional
digits for detailed classification. Classifying all products by HS code is possible first,
and all products worldwide are subject to tariffs in the basis of HS code. That is, the
products traded in the world have HS codes.
The “Smart Risk Pre-review System” learns the rules by Machine-learning in
each country’s HS. Machine learning is designed to learn by the rules and to find HS
codes of the products that are subject to.
The “Smart Risk Pre-review System” in this paper designs the Machine Learning
API with the product information provided by the Smart Risk Preview System, HS
code is found. To the Smart Risk Pre-review System to find HS codes, the Machine
Learning API marked the documents to be learned in Fig. 4. Documents are the func-
tion and usage of the product, information of the target of use and the manufacturer’s
company, and information of the importer. Documents are classified and stored on a
Data Base and used as Cloud Computing.
The product is defined by the product name and the product use manual in advance
review application. In addition, the HS code of the manufacturer and importer shall
be found in the importing country’s position in accordance with the rules of HS
code of the manufacturer and importer with the information of the manufacturer and
importer. At the actual work site, customs officials are finding HS codes, feeling very
difficult.

01- Handy Steamer-Preliminary Review Application HWP File

02- Handy Steamer-Importer Information PDF File
02- Handy Steamer-Manufacturer's Information PDF file
03- Handy Steam Iron-User Manuals Microsoft Word Documents
03- Handy Steam Iron-Product Specification Microsoft Word Document

Fig. 4 Example of an input document—HS code for smart risk pre-review system
160 Y. Oh and D. Park

In the era of super-connections of fourth industrial technologies, products become

more complex, and customs agents who find HS codes are also designing machine-
learning APIs to find HS codes in this paper to solve difficulties.

3.2 Deep Learning AI Design to Search for Technical

Regulations and Standards

The product’s HS code is determined by the AI engine designed in “Smart Risk Pre-
review System”. This system analyzes big data in cloud computing to find the correct
code. The HS code for this paper sample is “8516.40-0000”. Deep Learning AI design
is designed to establish technical regulations and standards. Analyze big data from the
data base stored in the cloud. Related laws are extracted from data mining from big
data analysis. The contents of protecting the technical characteristics and regulations
and standards of the product from HS code are designed with Deep Learning AI.
Through HS code secured by Deep Learning AI, it is found that four related statutes
may be involved as input values for electric iron “8516.40-000000”.
• Electrical Appliances and Consumer Products Safety Control Act
• Radio Waves Act
• Act on the Resource Circulation of Electrical and Electronic Equipment and
Vehicles
• Act on the Promotion of Saving and Recycling of Resources
However, the four technical provisions may or may not apply depending on the
technical characteristics of the electric iron and the information of the importer.
To establish technical regulatory requirements, the “Smart Risk Pre-review System”
may search for and confirm related information in big data in the clouding computing.
First, it is designed to add circuit diagrams First, it is designed to add circuit
diagrams, assembly drawings, and parts list as input information of the “Smart Risk
Pre-review System” as Fig. 5.

Fig. 5 Example of an input document for the “smart risk pre-review system”—technical regulation
Design of Artificial Intelligence for Smart Risk … 161

The “Smart Risk Pre-review System” uses AI Deep Learning API The import
history of importers, parts list, assembly drawings and circuit diagrams of the Handis
Timber, among the four statutes, including those subject to AI Deep Learning vari-
ables, concluded that the applicable laws were the “wave law” and the “safety
management of electrical and household goods. “The “Smart Risk Pre-review
System” utilizes the AI Deep Learning API and concludes that the standard under
the “transmission method” is equivalent to “KC-compliant registration” and that the
test criteria are KN14-1, KN14-2. However, reaching a conclusion on the “Electrical
and Household Goods Safety Management Act” requires more cloud and database
information. Deep learning is carried out with data on the manufacturer’s produc-
tion history, quality system information, certification of applied parts and safety
as AI Deep Learning variables. The required information will be provided in the
cloud or in government or certification authority big data using data mining as an
AI deep learning variable (feature). The more search learning about components in
the “Smart Risk Pre-review System”, the more advanced AI deep-learning variables
will be added, and the more accurate AI deep-learning performance will be carried
out on technical regulations and standards.

3.3 Designing Machine Learning AI for Conformity Check

and Quality Labelling Document

The “Smart Risk Pre-review System” analyzes reports and certificates for prod-
ucts. It is going to design AI machine-learning algorithms to check if products
were evaluated for suitability. It is designed to analyze big data in the cloud to
verify reports and certificates, or to analyze and validate job-input documents. The
product’s model name, certification number, manufacturer, importer, and vendor
information are provided as AI machine running variables and designed for anal-
ysis. AI Machine Learning search is a process to verify that compliance work has
already been completed. It is required to verify compliance by entering a test report
or certificate. Designs are also intended to display relevant technical regulations
after checking compliance. The technical regulations legislate items that require the
product to disclose to consumers. Technical regulation items are designed to provide
documentation displays automatically by the “Smart Risk Pre-review Systems” based
on the content of reports and certificates. The input data required for AI to analyze
the task is shown in Fig. 6.
162 Y. Oh and D. Park

Fig. 6 Example of the input document of the smart risk pre-review system—validity check

4 Conclusion

In this paper, we studied a “smart risk preliminary review system” that can provide
necessary information for market entry in response to non-tariff barriers. This system
evaluates the suitability of products and analyzes big data.
The function of this system is to determine the suitability of the product and to
analyze a Big Data to provide a solution. Because it is operated by cloud computing,
it performs all functions non-face-to-face.
It is designed to connect the information on the production, manufacture, distri-
bution, and customs of trade products using block chain. The information generated
by the” Smart Risk Pre-Review System” analyzes big data in a database stored in
the cloud. With AI relearning and AI machine learning, trade product information is
designed to be transported and operated in cloud computing networks.
The “Smart Risk Pre-Review System” applies to product certification, environ-
mental certification, hazard certification, and quality system certification. The HS
code search for trade product sales requirements uses AI machine learning, and the
technical rule search and standard search are designed to use AI deep learning algo-
rithms. In addition, the conformance check AI machine learning will be learned, and
the conformance DB will be designed to apply the labeling rules.
Design of Artificial Intelligence for Smart Risk … 163

References

1. Kim BW, Kim JH (2019) Progress and impact of US-China trade friction. KIET Ind Econ Rev
Spec 1–3
2. Yuri D, Erdal O, Choi MK (2019) Cybersecurity: attack and defense strategy, 230–231
3. Nakamoto S (2008) Bitcoin: a peer-to-peer electronic cash system. Available: http://bitcoin.
org/bitcoin.pdf
4. Son BH, Kim JH, Choi DH (2017) Major science and technology innovation policy tasks for
responding to the fourth industrial revolution. KISTER Issue Paper. Available: https://www.
kistep.re.kr/c3/sub3.jsp?brdType=R&bbIdx=11289
5. Ahn SH, Han KH, Ahn SS (2002) Internet Key exchange based on PKI in wireless environment.
Korean Inst Inform Sci Eng 29(2III):139–141
6. Lim JH (2018) Characteristics of blockchain technology, industry utilization and market
forecast. Daehan Assoc Bus Adm 21
7. Lee JH, Youn SW (2019) Study on smart contract in blockchain. Korean Inst Inform Sci Eng
2091–2093
8. Jee EK, Jung YL, Yoo JH, Bae DH (2019) Simulation and formal verification of blockchain
consensus algorithms: a case study and analysis. Commun Korean Inst Inform Sci Eng
37(4):37–48
9. Park CH, Kim MK, Kim HW (2018) A Study on building blockchain network and decentralized
application development based on Hyperledger Fabric. In: Proceedings of symposium of the
Korean institute of communications and information sciences, pp 5–6
10. Ministry of science (2017) ICT and future planning and ministry of justice. Electron Doc
Commentary 9
11. National IT industry promotion agency (2011) Electronic document security management
guidelines research report, pp 13–14
AI Analysis of Illegal Parking Data
at Seocho City

Donghyun Lim and Deawoo Park

Abstract The CCTV surveillance center (below to “Center”) in Seocho City Office
operates 3724 CCTVs. CCTV control is operating infrastructure such as 40 Gbps
self-fiber-connected network dedicated information and communication, 7 PB SAN
storage and 200 virtual machines to ensure the safety of citizens. Center grow up
the newest technique from now on. However, in view of Bia Data for illegal parking
and stopping judgment in Center relies on the old automatic software. In this paper,
we characterize images by local resident, neighborhood, site-specific, site-specific-
direction, day of week, hour of day and analysis of Bia Data for judging illegal
parking. We design an AI machine learning system that links the vehicle’s number
recognition algorithm for illegal parking. Finally we design a advanced system that
analyzes the status of illegal parking and stopping judgment in Seocho City Office,
where Big Data and AI are connected using spatial information and AI.

Keywords CCTV · Deep learning · Video analysis · Illegal parking · Parking stop
judgement

1 Introduction

The 4th Industrial Revolution technology is being applied to the real life of citizens.
5G smartphones receive information on the movement of patients with confirmed
COVID-19 in Seocho City. 5G smartphone users can receive information from
COVID-19 diagnostic tests, as well as phone connections with real-time map search
and AI voice recognition and voice commands to avoid infection routes. While

D. Lim · D. Park (B)

Department of Convergence Engineering, Hoseo Graduate School of Venture, Seoul, Korea
e-mail: [email protected]
D. Lim
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 165
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_13
166 D. Lim and D. Park

driving a vehicle, 5G smartphones take into account GPS and road traffic as well as
video calls, and AI searches for the optimal vehicle driving route with a map and
guides the vehicle with voice.
It is the traffic problem and garbage problem that Seocho City receives a lot
of civil complaints for urban management. In traffic problems, especially illegal
parking, which obstructs vehicle flow, is the most common complaint. To solve these
traffic problems, Seocho City installed parking control CCTVs and expanded and is
operating 348 parking control CCTVs in August 2020.
However, in the process of improving citizens’ life satisfaction and solving illegal
parking, parking enforcement CCTV has not disclosed the enforcement process data.
In addition, even after the illegal parking crackdown, only individual crackdowns are
notified. Only simple information such as installation location is open on the website
of Seoul and Seocho City as public parking control information [1, 2].
It is necessary to utilize the advanced technologies of the 4th Industrial Revolution
to improve the efficiency of the CCTV operation for parking control in Seocho City.
Parking control CCTV, introduced in Seoul city and autonomous districts since the
late 2000s, has introduced some artificial intelligence machine learning supervised
learning concepts, but in order to satisfy the actual parking control problem, the
introduction of artificial intelligence deep learning is necessary. In other words, when
determining the number of cases of parking enforcement CCTV analysis, reading,
and administrative disposition, artificial intelligence designed a convolutional neural
network of deep learning and an artificial intelligence algorithm for detection and
optimal judgment to analyze actual parking enforcement CCTV and illegal parking.
It should be applied to the crackdown.
AI-based vehicle discovery and vehicle number acquisition research are in
progress like Vehicle license plate area detection using artificial intelligence
deep learning [3], license plate recognition using polynomial-based RBFNNs
[4], vehicle number recognition using data expansion and CNN algorithm [5].
The parking control method using artificial intelligence deep learning is a ReID
technology for tracking criminal vehicles [6, 7] can be used. It also fits the govern-
ment’s policy direction, which understands the 4th Industrial Revolution technology
as improving administrative services to the public [8].
In this paper, after examining the actual parking control business process of Seocho
City Office, we apply a binary machine learning method based on artificial intelli-
gence supervised learning in the analysis and operation method of illegal parking
control CCTV. In big data analysis, information on the parking control process in
Seocho City is analyzed from the perspective of dong, time, day and system resources.
In addition, for the efficiency and improvement of illegal parking enforcement in
Seocho City Office, a parking enforcement method using artificial intelligence deep
learning is designed and proposed.
AI Analysis of Illegal Parking Data at Seocho City 167

2 Vehicle Number Recognition System Analysis

2.1 Data Preprocessing for Seocho City Vehicle Number

Recognition and Extraction

In the case of vehicle number recognition using a contrast difference, the distance
between the camera and the license plate, weather and shadows are greatly affected.
So, in most cases, it is used the region binarization [9] and the morphology tech-
nique(open and close operation) [9]. Through this pre-processing, image quality
improvement, lighting correction, and shadow distortion can be compensated.

2.1.1 Locally Adaptive Thresholding Method

Images captured in the field with a CCTV imaging device (CMOS) are difficult
to identify images due to various lighting interferences. Therefore, the binarization
technique is used to cancel unwanted illumination interference by using the contrast
of the monochrome image.
There are many types of binarization. There is the simplest binarization method,
global fixed thresholding method that binarizes based on a threshold value, but it is
not easy to specify a license plate or vehicle number using this method alone. This is
because the binarization success rate is high where the environment is prepared, but
the possibility of binarization failure is high outdoors where there is a lot of lighting
interference. Another method is Locally adaptive thresholding, which is used for
calculation by subtracting a constant from the brightness average for each pixel of
the image and calculating a threshold value like Fig. 1. This method is common [4].

Fig. 1 Locally adaptive thresholding (Left) and global fixed thresholding (Right)
168 D. Lim and D. Park

Fig. 2 Result of opening plus closing calculation [Before (Left), After (Right)]

2.1.2 Image Morphology Technique

Since the binarized image is noisy, it is impossible to process it without correction.

The method used at this time is the morphology technique. This technique maintains
the characteristics of the image and processes only the shape change. Morphology is
a concept similar to a mask and refers to a method of selecting pixels in an image in
various shapes. Expansion, erosion, closing, and opening are possible as a method
of calculating this, but generally in Fig. 2, opening and closing operations (Closing)
removes noise and spots [4].

2.1.3 License Plate and Number Extraction

The outline of the binarized image in the previous section is displayed using find-
contour among various functions of OpenCV. At this time, the license plate is
determined using the unique ratio index of the license plate, and the number is then
extracted by reflecting the ratio. If the ratio of the extraction number is not appro-
priate, the previous process is repeated to continue the process of finding the right
license plate and number. Figures 3 and 4 is a picture of license plate and number
extraction using findcontour function [4]. Various technologies have been devel-
oped to remove license plates and number distortion according to the shooting angle
[10].
AI Analysis of Illegal Parking Data at Seocho City 169

Fig. 3 Plate detection using by findcontour function [Outline formation (Left) and Plate box
formation (Right)]

Fig. 4 Number detection using by findcontour function [Outline formation (Left) and Number
box formation (Right)]

3 AI Machine Learning Analysis on Illegal Parking

in Seocho City

3.1 Process Analysis for Illegal Vehicle Number Recognition

AI machine learning supervised learning expresses the process of recognizing the

number of illegally parked vehicles in Seocho City. In order to increase the efficiency
of parking enforcement, the performance of each process for recognizing the vehicle
number is important. Only by improving the individual performance of each process
can be expected to improve the license plate recognition rate.
Seocho City parking enforcement work process is shown in Fig. 5. Parking control
CCTV attempts to recognize illegal parking vehicle license plates based on the
same principle as AI machine learning supervised learning according to the already
entered schedule. The CCTV control center manager visually checks the license plate
information for illegally parked vehicles that were automatically cracked down the
day before, deletes personal information, corrects errors, confirms the enforcement
information, and issues a notice.
170 D. Lim and D. Park

Fig. 5 Illegal parking work process

• SW analysis design for AI machine learning

The parking control program installed on the Seocho City parking control host
described in Fig. 7 checks the pre-registered schedule, checks whether parking is
available date/time, and performs a preprocessing process of binarizing by control-
ling the camera. After finding the feature point, extracting the license plate and
vehicle number in order, record it as a target for enforcement, and take pictures in
wide view and narrow view (this is called 1st capture). It moves to the next preset
defined in the setting to perform the vehicle detection process, and after a certain
time elapses after the first site crackdown, it returns to the memorized angle of view
and re-shoots to confirm the crackdown (this is called 2nd capture).
The currently used technology uses OpenCV-based supervised learning engine,
Fig. 6, a more advanced AI machine learning algorithm is designed to operate as an
AI machine learning supervised learning process.

3.2 AI Machine Learning for Illegal Parking Enforcement

Table 1 applied to Seocho City parking enforcement 348 CCTVs carried out 869,913
cases and 246,905 cases of first and second crackdowns for one year, of which only
10.427% of the first crackdown standards were actually fined.
• Learning AI machine learning by dong in Seocho City

There are 18 dongs in Seocho City. Figure 7 represents the number of CCTVs
installed in each building and the number of crackdowns in each building. It can
be seen that the number of CCTVs is low in places with large-scale apartment
AI Analysis of Illegal Parking Data at Seocho City 171

Fig. 6 Illegal parking software process

Table 1 Operating result review

Category CCTV Amount after Amount after Amount of Amount of
amount 1st detecting 2nd detecting fault judgment
Total 348 869,913 246,905 156,207 90,705
Rate − − 28.383% 17.957% 10.427%
(Upper/1st)
Rate − − − 63.266% 36.737%
(Upper/2nd)
172 D. Lim and D. Park

Fig. 7 Each step’s amount of illegal parking detection in 18 small province

complexes, and the number of CCTVs is high in places where shopping centers
and offices are concentrated. It can be seen that the number of crackdowns is high in
Seocho1-dong, Seocho2-dong, Seocho3-dong, Seocho4-dong, where there are many
floating populations such as Gangnam Station and Express Terminal, and Banpo4-
dong and Jamwon-dong. The peculiar point is that in Banpobon-dong, the amount
of enforcement is small, but the detection error rate is low.
Figure 8 expresses the value obtained by dividing the number of crackdowns per

Fig. 8 Result of [Amount of 1st/Amount of CCTV] in each small province

AI Analysis of Illegal Parking Data at Seocho City 173

Fig. 9 Illegal parking amount comparison of daily timeline

dong by the number of cameras per dong based on the first crackdown result. You
can see where the probability of finding a parked vehicle is high.
• AI machine learning results by time

The parking enforcement time in Seocho City is from 08:00 to 24:00. From 12 o’clock
to 14 o’clock, it is a grace time according to the meal time, and illegal parking is
not regulated except for a few places that are very fatal to traffic Fig. 9. Crackdowns
occur during the most active hours before and after lunch.
• AI machine learning results by day of the week

Comparison of the number of parking enforcement by day of the week Fig. 10 shows
that there is a strong tendency for illegal parking on Monday morning, Tuesday
afternoon, and Thursday evening. Comparison of the number of parking enforcement
on weekends Fig. 11 shows that there is a higher possibility of causing traffic jams
due to illegal parking on Saturdays than on weekdays, and less than 30% of illegal
parking on Sundays than on weekdays.
174 D. Lim and D. Park

Fig. 10 Illegal parking amount comparison of week

Fig. 11 Illegal parking amount comparison of sun, mon and sat

AI Analysis of Illegal Parking Data at Seocho City 175

4 AI Deep Learning Design to Improve Illegal Parking

Enforcement Efficiency

4.1 AI Deep Learning System Design

It analyze and design AI deep learning algorithms and methods to make illegal
parking enforcement and administrative disposition more efficient. For the maximum
use of AI deep learning server resources, all servers have to adjust to virtualization
based on hypervisor and operates thin provisioning.

4.2 AI Deep Learning Process Improvement Design

The current method of analyzing illegal parking enforcement video information and
operating some AI supervised learning machine learning-based engines has many
inefficient factors such as waste in process management and raising the question
of administrative measures for enforcement. Illegal parking enforcement schedule
is managed by AI supervised learning, but illegal parking enforcement vehicle and
vehicle number recognition is Fig. 12, it is suggested to design with AI deep learning.

4.3 AI Deep Learning Algorithm Application Design

Supervised learning machine learning-based algorithm, which is a method of illegal

parking control system, is shown in Fig. 13, the AI deep learning algorithm DNN
(Deep Neural Network) is converted to improve the performance of the entire illegal
parking process.
In AI Deep Learning DNN Algorithm, HL (Hidden Layer) 1 inputs illegal parking
status, vehicle number recognition, elapsed time calculation for schedule, real-time
status information, etc. as main functions to operate AI deep learning system and
illegal parking control system Improves performance.

4.4 Improvement Effect When Applying AI Deep Learning

When AI deep learning and DNN algorithm are applied to illegal parking enforce-
ment, improvement effects as shown in Table 2 can be obtained. Through AI deep
learning and DNN algorithm, it can be applied to buildings where illegal parking
enforcement and vehicle number recognition efficiency is low, thereby improving
parking enforcement efficiency. In addition, by operating the deep learning DNN
algorithm according to the number of cases in the field, the administrative power
176 D. Lim and D. Park

Fig. 12 Improvement design for AI deep learning SW process

Fig. 13 Improvement design

for AI deep learning process

of enforcement and prevention can be concentrated in the days and times of high
frequency of illegally parked vehicles. The final goal, the 4th Industrial Revolu-
tion technology, can be reflected in administrative policy to increase the safety and
satisfaction of citizens.
AI Analysis of Illegal Parking Data at Seocho City 177

Table 2 Comparison of AI machine learning versus AI deep learning

Category AI machine learning (ML) AI deep learning Improvement effect
(DL)
License plate 100 ch × 3 frames per second ML × 4 times Infrastructure
recognition (Parallel efficiency
discrimination processing) improvement
amount
Number of license Manual learning in 1–2 DL algorithm Labor cost
plate recognition times/month self-learning reduction and
learning times improved accuracy
Illegal parking Administrator visually checks No intervention of Administrative
control judgment and enters the ML high reliability power reduction by
method algorithm by preemptive
managers exception handling
and labor cost
reduction

5 Conclusion

When operating illegal parking control CCTV, some methods of detecting vehi-
cles and recognizing vehicle numbers are operated based on AI machine learning
supervised learning. However, inefficient SW/HW operation, location/time/day of
the week, etc., the classification and regression methods of machine learning super-
vised learning are not effectively reflected. Therefore, there is a problem in that a
lot of effort and time are consumed by the administrator due to the uniform illegal
parking control method.
In this paper, we presented big data analysis by reflecting situations such as
SW/HW operation and winter/time/day of the week through classification and regres-
sion methods of AI machine learning supervised learning. In addition, AI deep
learning DNN algorithm was designed based on the AI learning method results
by analyzing AI deep learning in order to efficiently control illegal parking vehi-
cles. In addition, in HL (Hidden Layer) 1 using the DNN algorithm, the AI deep
learning system is operated by inputting illegal parking status, vehicle number recog-
nition, elapsed time calculation for schedule, real-time status information, etc. as
main functions. An improvement plan was designed to improve performance. By
reflecting the AI deep learning system design, AI deep learning process improve-
ment design, and AI deep learning algorithm applied design, the amount of illegal
license plate recognition is increased by 4 times, the number of license plate recog-
nition learning is reduced, the reliability of illegal parking enforcement judgment is
improved, and real-time data mining The improvement effect on the possibility of
report was confirmed.
In the future, research on algorithms that can be applied to more than 10 illegal
parking control deep learning process variables and research that can apply deep
learning to disaster situations and crime prevention using illegal parking control
CCTV is needed.
178 D. Lim and D. Park

References

1. Seoul City parking violation control inquiry system. https://cartax.seoul.go.kr/

2. Seocho City parking enforcement system. https://traffic.seocho.seoul.kr/
3. Jeong Y, Ansari I, Shim J, Lee J (2017) A car plate area detection system using deep convolution
neural network. J Korea Multimedia Soc 20(8):1166–1174
4. Kim S, Oh S, Kim J (2015) Recognition of vehicle license plate using polynomial-based
RBFNNs. The Korean Institute of Electrical Engineers, pp 1361–1362
5. Lee G, Kim Y, Lee D, Kang H (2020) Vehicle number recognition using data extension and
CNN algorithm. Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography,
pp 237–240
6. Lee S, Chu S, Kwon K, Cho N (2017) Pedestrian detection and re-identification for intelligent
CCTV systems. J Korean Inst Commun Sci 34(7):40–47
7. Chae S, Choi H, Kim I (2018) Person re-identification in multiple camera environment. Inst
Electron Inform Eng 380–381
8. Ministry of the Interior and Safety; Public Service, Fly with A Digital Technology, Ministry
of the Interior and Safety, pp 180–186
9. Kim D, Moon H (2019) Implementation of pre-and-post processing algorithm to improve LPR
(License plate recognition). Trans Korean Inst Electr Eng 68(12):1594–1600
10. Kim J (2011) Distortion invariant vehicle license plate extraction and recognition algorithm.
Korea Contents Assoc 11(3):113–120
Regularized Categorical Embedding
for Effective Demand Forecasting of Bike
Sharing System

Sangho Ahn, Hansol Ko, and Juyoung Kang

Abstract The value of sharing economy services is increasing every year, and
demand forecasting based service operations are essential for sustainable growth.
For effective demand forecasting, this study proposes a categorical embedding based
neural network model. The performance of this model is better than the traditional
one-hot encoding based prediction; however, there are difficulties in creating a gener-
alized prediction model due to the possibility of over-fitting of training data. Accord-
ingly, it is possible to predict optimal demand by showing regularized performance
applying techniques such as Batch Normalization, Dropout, and Cyclical Learning
to the neural network. This methodology is applied to the Bike Sharing System to
forecast bicycle rental demand by stations. In addition, in order to use the character-
istics of global learning categories, uniform manifold approximation and projection
(UMAP)-based dimensionality reduction technique is performed on the embeddings.
The dimension-reduced embeddings are projected on the coordinate plane and used
for K-means based cluster analysis, thereby providing an effective analysis result for
demand patterns.

Keywords Sharing economy · Bike sharing system · Demand forecasting ·

Categorical embedding · Uniform manifold approximation and projection (UMAP)

S. Ahn · H. Ko · J. Kang (B)

Department of e-Business, Ajou University, Suwon, Republic of Korea
e-mail: [email protected]
S. Ahn
e-mail: [email protected]
H. Ko
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 179
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_14
180 S. Ahn et al.

1 Introduction

Over the past few years, the emergence and development of a sharing economy
platform has transformed the way people “share” a single public good, beyond the
conventional “rental” concept of lending private property; as a result, shared items
have appeared in various markets. According to PwC, a global survey agency, the
sharing economy market is expected to grow from $15 billion in 2014 to about
$350 billion in 2025. However, since economic services are based on sharing limited
resources, optimization of service-demand by intelligent systems is a strategic task
[1]. Specifically, the example of Bike Sharing System involves active rentals with
several rental stations. In order to ensure that the service operates smoothly, it is
necessary to analyze the bicycle rental pattern and model it so that rental stations use
it to relocate bicycles. In fact, Fig. 1, based on about 9000 opinions on the citizen
opinion bulletin board, confirms that there were complaints of relocation in Seoul’s
bike sharing system.
On the other hand, demand forecasting can lead to robust results by applying a
traditional regression or tree-based model to tabular data comprising a combination
of continuous and categorical variables. These methodologies have the advantage
of determining feature importance, so that it is possible to find out which variables
have a significant influence on demand and which variables are unnecessary [2, 3].
However, since the categorical variable was converted into a one-hot encoder to

Fig. 1 Wordcloud of the bulletin board from Seoul citizens’opinion

Regularized Categorical Embedding for Effective Demand … 181

generate a model, the relationship between the categories could not be confirmed,
and it was not suitable for explaining similar rental patterns. In addition, in the bike
sharing system, which has a different distribution of rental demand for each rental
station, it is necessary to use multiple forecasting models for each rental station rather
than a universal model; thus, the entire rental pattern cannot be implicated in each
model.
Therefore, in order to solve these problems while meeting the task of forecasting
demand by rental stations, this study proposes a neural network-based demand fore-
casting methodology applying categorical embedding for the bike sharing system.
Specifically, it uses not only continuous variables such as climate information but
also categorical embedding variables such as time information to train a neural
network for each rental station. The study method shows that this is an excellent
technique for forecasting rental demand. In addition, cluster analysis is performed
to classify similar rental stations by embedding rental station information using the
same methodology extensively. Unlike existing traditional practices, it will enrich
demand forecast that is insufficient for each rental model by using rental embed-
dings and clusters that implicate the global trend of total rental data. Moreover, since
multiple regularization techniques are used in the model generation and training
stage, generalized results can be expected without overfitting.
The overall study contribution is as follows.
(1) We propose a regularized categorical embedding (RCE) methodology for effec-
tive embedding of categorical variables and generalized prediction perfor-
mance.
(2) It shows that the model’s prediction performance is superior to others, which
are recognized to have excellent performance in the existing tabular data.
(3) Categorical embedding, dimensionality reduction, and clustering methodolo-
gies are combined to find similar demand patterns.
(4) Effective relocation is possible in shared economic services, so it is possible to
provide effective services to users.

2 Related Works

2.1 Demand Forecasting for Bike Sharing Systems

Over the past few years, research on predicting demand for shared bicycle rental
services has been ongoing. First, attempts were made to add several variables for
prediction. Reference [4] generated regression equations using “gu” variables (which
are administrative areas) to predict bicycle demand for specific regions. According
to [5], most existing studies on demand forecasting have overlooked the variables
of weather change but suggested that these variables significantly influence demand
forecasting. Reference [6] conducted time series analysis using the Holt-Winters
method and performed demand forecasting.
182 S. Ahn et al.

There are trends to apply various machine learning techniques. For instance, [7]
used various techniques such as Ridge linear regression, support vector regression
(SVR), random forest, and gradient boosted trees based on shared bicycle rental data
in 2011 and 2012 in Washington, D.C. Similarly, [8] used Recurrent Neural Networks
(RNN) to solve imbalanced placement of shared bicycles by rental locations and
conducted research on evenly relocating rental bicycles based on shared bicycle data
in New York City. Among them, a remarkable study is by [9], which used a station-
centric model to predict the rental demand of bike sharing systems. In addition, a
methodology was used to increase the performance by extracting global features and
clustering adjacent rental stations by creating a City-Centric model that aggregates
the entire station data. In this study, we adopted the city-centric technique to implicate
the global pattern.
However, there is a limitation that such a study on bicycle sharing does not
completely solve the problem of demand imbalances by rental stores. In most studies,
it was difficult to expect accurate demand predictions to meet the latest trends using
only user rental data for about a year or so. In addition, existing studies that performed
daily and monthly analysis, not time-based analysis, could not determine when
time zone relocations should occur. Finally, there was a limitation that the one-
hot encoding method for categorical variables could not be used to reveal the deep
relationships between categorical variables. Therefore, this study not only uses three-
year data but also categorical variables for detailed time zones; therefore, it is possible
to predict relocation by time zones. In addition, since it adopts a methodology for
embedding categorical variables, it will be possible to understand intuitively for each
variable and demonstrate improved performance.

2.2 Categorical Embedding

This study forecasts demand based on tabular data called public bicycle rental data,
and an embedding method is applied to categorical data. Research on embedding
has been actively conducted across various fields, especially in natural language
processing [10] and recommendation system [11]. In fact, in the field of text mining,
word embedding research is at the forefront. In word embedding research using
artificial neural networks, [12] introduces Word2Vec, which learns the probability
of a word appearing through the relationship between the surrounding context and
a specific word. Using this, [13] propose a Structural Deep Network Embedding
method to effectively detect nonlinear network structures and preserve global and
local structures. Also, [14] shows that Deep Embedded Clustering is a method for
simultaneously learning feature expressions and cluster allocations using deep neural
networks and shows that Unsupervised Deep Embedding is also used in the analysis.
As such, research is underway to embed information in various fields.
However, studies to solve the problem by applying deep learning to data including
categorical variables have not been actively conducted [15]. These attempts were
presented in two studies that performed well on Kaggle, which solved data analysis
Regularized Categorical Embedding for Effective Demand … 183

problems, and were highly regarded academically. In fact, [16] won the first place
by applying a deep learning model to categorical metadata in the ECML/PKDD
15: Taxi Trajectory Prediction (I)” competition, suggesting that embedding actually
affected performance. In the “Rossmann Store Sales” competition, [17] applied the
methodology of embedding categorical data to demand forecasting problems, and not
only secured third place but the categorical data represented in the vector dimension
also improved the prediction performance. This has high value as an example of
interpretability and transfer learning in that the meaning of the relationship between
the data in each category is expressed together. The above embedding case studies
show that academic and practical results in deep learning models can be applied to
table data by representing categorical data in vector dimensions; however, there were
limitations in overfitting because they did not deal with normalization. In addition, a
specific utilization plan for the relationship between learned embeddings of variables
have not been suggested.

3 Proposed Method

For the purpose of this study, two processes are required for effective demand fore-
casting and performance evaluation of the bike sharing system, as shown in Fig. 2.
The first step is to design and train a categorical embedding-based neural network,
or Regularized Categorical Embedding (RCE) model, which applies regularization
techniques to predict demand by rental station.
The second step is to solve the problem that a suitable model cannot be found
in a specific rental station due to data imbalances by rental stations (Station), and
rental embedding is solved through cluster analysis applying Uniform Manifold
Approximation and Projection (UMAP) and K-means.

Fig. 2 Overview of our proposed model

184 S. Ahn et al.

Fig. 3 Our regularized categorical embedding network

3.1 Categorical Embedding Based Network

Demand forecasting model is a regression model that predicts continuous depen-

dent variables and must receive numerical data. Therefore, continuous variables can
be used without modification; however, categorical variables must be expressed in
numerical form to proceed with the analysis. There are two main methods to solve
this: (a) one-hot encoding based method, and (b) embedding-based method. From
an analytical perspective, embedding-based methodology is more effective as it has
better predictive performance and learns the data properties [18].
The RCE method was also used for embedding categorical variables, and various
normalization techniques were applied to create and distribute a practically useful
rental prediction model. First, batch normalization is used to normalize the distribu-
tion of different input values for each layer. It is a representative method to improve
speed, performance, and stability of the artificial neural network, and prevent over-
fitting during learning. [19]. Second, dropout is implemented by randomly excluding
the nodes of the layer at a specific rate for each training, and not only produces the
same effect as ensemble multiple models, but also prevents overfitting as in the
previous batch normalization [20]. By including these techniques, it is possible to
obtain a generalized prediction model without overfitting to the learning data, and to
obtain well-trained embedding. The overall model architecture is presented in Fig. 3.

3.2 Model Optimization

An optimization stage was implemented in the learning stage through a pre-processed

dataset, and a learning methodology termed “cyclical learning” was adopted. It is
Regularized Categorical Embedding for Effective Demand … 185

possible to reach the global minimum effectively even in small iterations by applying
a policy that covers both the basic learning rate and maximum learning rate, rather
than using a fixed learning rate [21]. When this methodology is used, first, a target
learning rate is required. When learning for the first time, it starts from a value that
divides the learning rate by a designated factor and gradually increases to the target
learning rate. After achieving the target learning rate, it gradually decreases to the
learning rate that progressed first time and concludes learning to complete one cycle.
The loss function used for optimization is a mean squared error (MSE) and the Adam
Optimizer is used by the optimizer.

3.3 Demand Forecast and Station Embedding

The above RCE methodology is applied to the data for each rental location to generate
a demand forecast model. First, the reason for creating a model for each rental
place, rather than a universal model based on the entire data, is that previous studies
confirmed that a different number of units and rental patterns exist for each bicycle
rental location. In order to confirm this, model generation for each rental shop is
completed and the evaluation indicators are displayed as a histogram to examine the
distribution.
The advantage of the RCE methodology is that even if a new categorical variable
is included, it does not add as many dimensions as one-hot encoding; besides, it is
easy to train the new categorical variable to reflect this tendency. Therefore, data
of the previous rental-specific models were merged, and the rental station itself was
added as a variable for embedding information in the rental station. Therefore, it
is possible to learn the tendency for overall rental demand, that is, the pattern, and
the entire rental shop learns this tendency and is represented by multi-dimensional
continuous variables.

3.4 Projection Embedding via UMAP and Clustering

Analysis

All learning categorical variables are represented by different multi-dimensional

embedding vectors, which preserves a wealth of information, but has the disadvantage
of being complex and difficult to interpret. Therefore, several techniques to visualize
these multi-dimensional representations using two-dimensional reduction techniques
were applied [22]. Initially, methods such as Principle Component Analysis (PCA)
or t-SNE were used in dimensionality reduction methods, but a UMAP method
that can compress multi-dimensional information more efficiently and describe the
global structure has been proposed. Indeed, [23] projected movie information or
user information acquired in the recommendation system in two dimensions using
186 S. Ahn et al.

UMAP and showed that this is more effective than other techniques. Therefore,
UMAP dimensionality reductions can be applied to the embedding vectors learned
through the RCE technique, and each one can be visualized and expressed.
In this study, UMAP was used to analyze the pattern by dimensionally reducing the
embedding vector that was learned earlier and clustering it. Based on this, it is possible
to effectively represent information on rental embedding vectors that learned global
trends based on the entire data in two dimensions to create clusters between rental
stations with similar rental patterns. Therefore, rental demand forecast, which was
inadequate by the rental model due to unbalanced data or patterns, is supplemented
with the model with highest predictive performance in the cluster, and the rental
pattern is analyzed.

4 Experiments

4.1 Datasets

In this study, about 12.5 million rental datasets from January 2017 to May 2019 were
collected from the “Seoul City Public Bike Usage Status” to predict the demand on
a specific “rental station” and analyze time pattern for bike relocation. At this point,
since there were a few different data types in the yearly dataset, they were all unified
in the same format, and preprocessing was performed for missing values and outliers.
In addition, precipitation, fine dust, and temperature information for each time zone
were collected using the Seoul Metropolitan Meteorological Administration API
to add climate information. Moreover, climate information was not subdivided into
administrative regions because it was confirmed in the EDA (Exploratory Data Anal-
ysis) process that there were no significant climatic differences in each region. By
combining the two types of data collected using this method, it was possible to
construct the final type of data necessary for demand forecasting. In this dataset,
categorical variables are time information such as rental time, rental day, rental
month, season, and weekend. Continuous variables are climate information such as
humidity, wind speed, temperature, precipitation, fine dust, and ultrafine dust.

4.2 Experimental Setup

This experiment was implemented through the Pytorch-based fast.ai library in Google
Colab’s GPU Tesla K80 environment. In order to minimize the mean squared error
(MSE) loss function, the Adam-Optimizer [24], which can reach global solutions
faster than other optimizers, was used. Various hyperparameters were used for
learning via the RCE methodology. Most of them used the default value of fast.ai,
but the learning speed was 0.04, and Cyclical Learning rates were applied to learn for
Regularized Categorical Embedding for Effective Demand … 187

Table 1 The result of

Variable # of Category # of dimension # of dimension
dimension with embedding
Name (Embedding) (One-hot)
and one-hot encoding
Station ID 1505 96 1505
Rental 24 9 24
hour
Rental 7 5 7
day
Rental 12 6 12
month
Rental 4 3 4
season
Rental 2 2 2
weekend

10 and 30 epochs. In the case of category variables, the size of the embedding must
be defined. In the case of fast.ai, the size of embedding N_e is basically set by the
following equation based on the number of categories N_o per original categorical
variable [12].

Ne = r ound(1.6 × No0.56 )

Then, the result of dimensions, embeddings, and one-hot encoding method, to

which the above formula is applied, are given in Table 1, respectively. It can be
observed that as the number of categories increase, the embedding method represents
data in a much smaller number of dimensions than one-hot encoding.

4.3 Evaluation Metrics

We tried to compare the results through R Square, which is one of the elements of
performance evaluation borrowed from the traditional statistical model. In addition,
to generalized performance evaluation, k-fold cross-validation was implemented.
This is to derive the average of the predicted values by applying the algorithm and
divide the data set by a certain number that can prevent overfitting [25]. In this study,
fivefold cross-validation was performed.

4.4 Baseline

First, to evaluate the performance of the embedding model, a predictive model was
created for four rental locations with a large number of rentals. Furthermore, for
comparison, one-hot encoding multiple linear regression, XGBoost, and random
188 S. Ahn et al.

forest models were compared, and the results are as follows. The performance of the
proposed RCE model is indicated by the results of 30 and 10 learning cycles. The
mentioned models are LR (Linear Regression), XGB (XGBoost), and RF (Random
Forest), respectively, and the results of 100 learning cycles are indicated.
(1) Linear Regression (LR): Linear regression is a widely used technique for
predicting dependent variables based on several independent variables.
(2) XGBoosting (XGB): XGB is an effective and widely used machine learning
method that extends beyond billions of examples using significantly fewer
resources than existing systems [26].
(3) Random Forest (RF): Classification and regression by randomForest, the best
split among all variables, is divided using the highest predicted value among
each predictor variable that is randomly selected by the node [27].

5 Results

5.1 Model Performance

Table 2 shows the performance of RCE models and those introduced in the baseline
for the four rental shops with high rental volume. The results show that the embedding
model has the best performance compared to one-hot encoding, as demonstrated
in several studies. In addition, compared to traditional models, the performance
of neural network-based models is found to be superior. The performance of the
proposed NN-based models was indicated by 15 learning cycles and 100 learning
cycles for LR, XGB, and RF.

Table 2 Evaluation results

NN+Embeding NN+One-hot LR+One-hot XGB+one-hot RF+One-hot
(RCE)
Hongik 0.5558 0.5463 0.4625 0.2975 0.2929
university
station exit
2(113)
Yeouinaru 0.6092 0.6068 0.3710 0.2904 0.2857
station exit 1
(207)
Ttukseom 0.6083 0.5993 0.4318 0.3215 0.3197
amusement park
station exit 1
(502)
Mapo-gu 0.5326 0.5270 0.3835 0.3562 0.3545
community
sports center
(152)
Regularized Categorical Embedding for Effective Demand … 189

As the results show, the performance of forecasting the RCE methodology with
embedding in categorical variables was the highest under the same conditions.
However, even the best-performing RCE model did not perform well in all rental-
specific models. Through this, we can see that R Square had a similar pattern as
regular distribution, and rental shops had difficulty in finding the pattern with only
rental models.

5.2 Station Embedding and Clustering

We compared the performance of the RCE methodology and the baseline at the top
four rental stations with the highest volume of rental history and found that the RCE
methodology was the best performer, but there were rental stations that were also
difficult to apply. Therefore, to solve this problem, we tried to embed the rental station
itself as a variable, and the result is illustrated in the following figure. This enables
the following simple distance formula to calculate the similarity between 1500 rental
stations.

distance = (ea1 − eb1 )2 + . . . + (ean − ebn )n

A bigger advantage is that cluster analysis can be performed by dimensioning

multidimensional embedding vectors. As mentioned earlier, this study has reduced
the 96 dimensions of the rental station to two dimensions by applying UMAP. The
reduced dimensions are known to be in the best form to describe the entire data
structure. Therefore, a cluster analysis using K-mean was performed based on the
reduced two dimensions. Figure 4 produced clusters with four rental patterns through

Fig. 4 Station embedding scatter plot colored by cluster using K-means

190 S. Ahn et al.

K-means, and models with R Square values below 0.3 can be used practically, such
as replacing them with models with values of R Square in clusters greater than 0.3.

5.3 Rental Station Pattern Analysis: Time Information

The first pattern in the rental cluster relates to time information. As shown in Fig. 4,
a group of six rental locations with well-characterized features in four clusters was
sampled to analyze the number and pattern of rental time variables, and the analysis
results are illustrated in Fig. 5.
The nine rental stations located in A-1 group observed distinctive patterns of time
zones. The existence of such a clear pattern confirms that a model with relatively good
predictive performance is generated. On the other hand, even if they belonged to the
same group A, the five rental stations located in A-2 group could not confirm the clear
pattern, but the remaining rental stations, except “Cheonggyecheon Yeongdogyo
Bridge,” had a small number of platforms. In common, it was found that it was
a rental station with very low number of rentals. Four rental stations belonging to
cluster B always had a flat rental, but it was confirmed that there were patterns of
more than 700 rentals at 8 am. This related to work time, and it can be assumed that
many office goers use public bicycles at this time of the day. Next, six rental stations,
belonging to group C-1 of cluster C, had rental patterns by time zone between the
rental centers except “Eunpyeong Arts Center” rental station. It was confirmed that
the average number of rental hours during the day was much higher than at other
rental locations. It was confirmed that five rental stations belonging to C-2 group had
a pattern similar to A-1 group, but an incomplete pattern. Finally, it can be confirmed

Fig. 5 Analysis of rental patterns over time, groups A-1, A-2, B from top left, and groups C-1,
C-2, D from bottom left
Regularized Categorical Embedding for Effective Demand … 191

Fig. 6 Analysis of rental patterns over temperature, groups A-1, A-2, B from top left, and groups
C-1, C-2, D from bottom left

that incomplete rental patterns, such as C-1 and C-2, exist in the four rental stations
belonging to cluster D.

5.4 Rental Station Pattern Analysis: Temperature

The second pattern in the rental community relates to temperature. Similar to the
previous analysis, six groups of rental stations were sampled from four clusters to
analyze the number of rental patterns and the pattern of temperature variables, and
the analysis results are shown in Fig. 6.
First, the nine rental shops belonging to A-1 group clearly showed a pattern with
the highest number of rentals when the temperature was about 25°. On the other
hand, the rental pattern was not clearly revealed in the four rental offices belonging
to A-2 group, which was confirmed because of the small number of rentals. The
four rental stations belonging to cluster B, similar to A-1 group, showed a clear
rental pattern according to the temperature, confirming that active rentals occurred
when the temperature was around 25°. Next, C-1 group and C-2 group revealed clear
patterns, similar to A-1 group and B group, which demonstrated good performance
unlike the previous time zone. However, it is confirmed that there were incomplete
patterns in the four rental stations belonging to D cluster.
192 S. Ahn et al.

6 Conclusion

In this study, we propose a regularized categorical embedding methodology that

can learn not only the station-centric model but also the overall trend to predict the
rental demand for shared bicycles. Our model shows better predictive performance
from a minimum of 16% to a maximum of 53% through embedding with non-
sparse properties, while having fewer dimensions for each category in the categorical
variables compared to one-hot encoding. Also, the rental station itself can be learned
by embedding and combining with a dimensional reduction method based on UMAP,
which can be effectively used in terms of analysis. This was an efficient way to
find similar rental locations through cluster analysis and was used to describe some
rental locations that were difficult to explain in the rental-oriented model. This can
be confirmed through the distribution of variables of rental stations belonging to
the same rental pattern cluster and EDA. The proposed methodology is expected to
be an effective prediction for station-centric models with unbalanced performance
distribution and the prediction of rental demand for shared bicycles.
However, this study has two limitations. First, the high performance RCE model
for each rental station needs to train a new model for each station, so the deeper the
layer and the larger the number of rental stations, the higher the cost of the neural
network learning. Second, the effectiveness of the cluster analysis was evaluated
qualitatively, without numerical proof, thereby limiting its precise use. Therefore,
such problems must be developed to reduce the computational cost by transferring the
learned embedding variable, and to prove the similarity by fitting the cluster-specific
pattern as a linear function.

Acknowledgements This research was supported by the MSIT (Ministry of Science and ICT),
Korea, under the ITRC (Information Technology Research Center) support program (IITP-2020-
2018-0-01424) supervised by the IITP (Institute for Information and communications Technology
Promotion).

References

1. Guo L et al (2018) Quick answer for big data in sharing economy: innovative computer archi-
tecture design facilitating optimal service-demand matching. IEEE Trans Automation Sci Eng
15(4):1494–1506
2. Lindsey C, Sheather S (2010) Variable selection in linear regression. Stata J 10(4):650–669
3. Zheng H, Yuan J, Chen L (2017) Short-term load forecasting using EMD-LSTM neural
networks with a Xgboost algorithm for feature importance evaluation. Energies 10(8):1168
4. Kim KJK, Choi K, Keechoo (2011) Development of regression-based bike direct demand
models. Korean Soc Civil Eng D 31(4D):489–496
5. Do M, Noh Y-S (2014) Analysis of the affecting factors on the bike-sharing demand focused
on Daejeon City. Korean Soc Civil Eng 34(5):1517–1524
6. Lim H, Chung K (2019) Development of demand forecasting model for seoul shared bicycle.
J Korea Contents Assoc 19(1):132–140
Regularized Categorical Embedding for Effective Demand … 193

7. Yin Y-C, Lee C-S, Wong Y-P (2012) Demand prediction of bicycle sharing systems.
URL http://cs229.stanford.edu/proj2014/YuchunYin,ChiShuenLee,Yu-PoWong,DemandPre
dictionofBicycleSharingSystems.pdf
8. Chen P-C et al (2020) Predicting station level demand in a bike-sharing system using recurrent
neural networks. IET Intell Transp Syst
9. Zeng M et al (2016) Improving demand prediction in bike sharing system by learning global
features. Mach Learn Large Scale Transp Syst (LSTS) @KDD-16
10. Howard J, Ruder S (2018) Universal language model fine-tuning for text classification. arXiv
preprint arXiv:1801.06146
11. Liu DC et al 2017 Related pins at pinterest: the evolution of a real-world recommender system.
In: Proceedings of the 26th international conference on world wide web companion
12. Mikolov T et al (2013) Distributed representations of words and phrases and their composi-
tionality. In: Advances in neural information processing systems
13. Wang D, Cui P, Zhu W (2016) Structural deep network embedding. In: Proceedings of the 22nd
ACM SIGKDD international conference on knowledge discovery and data mining
14. Xie J, Girshick R, Farhadi A (2016) Unsupervised deep embedding for clustering analysis. In:
International conference on machine learning
15. Howard J, Gugger S (2020) Fastai: a layered API for deep learning. Information 11(2):108
16. De Brébisson A et al (2015) Artificial neural networks applied to taxi destination prediction.
arXiv preprint arXiv:1508.00021
17. Guo C, Berkhahn F (2016) Entity embeddings of categorical variables. arXiv preprint arXiv:
1604.06737
18. Zhang K et al (2015) From categorical to numerical: multiple transitive distance learning and
embedding. In: Proceedings of the 2015 SIAM international conference on data mining. SIAM
19. Ioffe S, Szegedy C (2015) Batch normalization: accelerating deep network training by reducing
internal covariate shift. arXiv preprint arXiv:1502.03167
20. Srivastava N et al (2014) Dropout: a simple way to prevent neural networks from overfitting.
J Mach Learn Res 15(1):1929–1958
21. Smith LN (2017) Cyclical learning rates for training neural networks. In: 2017 IEEE winter
conference on applications of computer vision (WACV). IEEE
22. McInnes L, Healy J, Melville J (2018) Umap: uniform manifold approximation and projection
for dimension reduction. arXiv preprint arXiv:1802.03426
23. Kang K et al (2019) Recommender system using sequential and global preference via attention
mechanism and topic modeling. In: Proceedings of the 28th ACM international conference on
information and knowledge management
24. Kingma DP, Ba J (2014) Adam: a method for stochastic optimization. arXiv preprint arXiv:
1412.6980
25. Fushiki T (2011) Estimation of prediction error by using K-fold cross-validation. Stat Comput
21(2):137–146
26. Chen T, Guestrin C (2016) Xgboost: a scalable tree boosting system. In: Proceedings of the
22nd ACM SIGKDD international conference on knowledge discovery and data mining
27. Kam HT (1995) Random decision forest. In: Proceedings of the 3rd international conference
on document analysis and recognition, Montreal, Canada
Development of a Model for Predicting
the Demand for Bilingual Teachers
in Elementary Schools to Support
Multicultural Families—Based on NEIS
Data

Jinmyung Choi and Dooyeon Kim

Abstract Education-related administrative agencies intend to analyze NEIS data

that has been in operation for over 15 years by applying big data and artificial intel-
ligence technologies to predict social issues and establish policies. Therefore, this
study developed a test model using big data to analyze the data of school sites and data
held by administrative agencies and applied the actual data to this model to analyze
the policy’s appropriateness and effectiveness. From 2013 to 2018, a learning model
was created based on the number of bilingual teachers for elementary school students
in multicultural families in Gyeongsangnam-do. As a result of predicting the number
of bilingual teachers in 2015, there were 40, but the test model was analyzed to require
42. It was confirmed that the test model’s accuracy could be further improved if more
data on the learning model is accumulated. This study’s results can help establish
and implement policies in various educational fields through big data analysis.

Keywords Bilingual · Education · Multicultural · Analytical model · Big data ·

Machine learning

1 Introduction

As defined by dictionary, education is a process of teaching, training, and learning,

especially in schools, colleges, or universities, to improve knowledge and develop
skills [1]. According to the dictionary’s definition of education, people can learn a lot
through education in school and gain abilities. However, it was found that children
of multicultural families in Korea have difficulty learning or show relatively lower
academic achievement than children of typical families [2]. There may be several
reasons for this, but the difference in speaking ability and understanding of a language

J. Choi · D. Kim (B)

Department of Convergence IT, Konyang University, Chungchungnam-Do, Republic of Korea
e-mail: [email protected]
J. Choi
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 195
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_15
196 J. Choi and D. Kim

can have a significant influence [2]. Also, the role of a teacher with the ability to
understand the cultures of different communities to remove cultural boundaries, to
understand different issues to communicate correctly with children of multicultural
families, and to increase academic success is vital [3, 4]. Thus, it is necessary to have
a bilingual teacher with qualifications verified through sufficient training. However,
it is difficult to determine whether the number of bilingual teachers is sufficiently
secured at school sites.
Recently, intelligent services are emerging as new value creation systems, such
as expanding user convenience with information services utilizing big data analysis
and artificial intelligence and predicting future situations. Using these technologies,
we develop a model that predicts the demand for bilingual teachers necessary for
the smooth education of children from multicultural families. To develop a bilin-
gual teacher demand model, data of bilingual teachers from the Gyeongsangnam-
do Office of Education were used from NEIS, which is being used in educational
administration.
Chapter 2 of this paper analyzes the learning status of children of multicultural
families, explains the characteristics of NEIS, and explains the method and procedure
for analyzing big data. Chapter 3 explains the process of developing a big data analysis
model, and Chap. 4 explains the suitability of the analysis model developed in this
study by comparing the results of analyzing the demand for bilingual teachers using
an analysis model and the number of bilingual teachers actually employed in schools.

2 Related Research and Status

2.1 Academic Achievement Status of Children

of Multicultural Families

Table 1 shows the current status of children from multicultural families in Korea from
2014 to 2018, and as of 2018, the number of children from multicultural families in
Gyeongsangnam-do is 17,723 [5].

Table 1 Current status of children of multicultural families in the Republic of Korea by age
Year Status by age (Persons)
Total Under 6 years old 7–12 years old 13–15 years old 16–18 years old
2018 237,506 114,125 92,368 19,164 11,849
2017 222,455 115,085 81,826 15,753 9,791
2016 201,333 113,506 56,768 17,453 13,606
2015 197,550 116,068 61,625 12,567 7,290
2014 204,204 121,310 49,929 19,499 13,466
Development of a Model for Predicting the Demand … 197

In the study of the learning status of children of multicultural families [2], three
sub-areas including learning motivation, learning strategy, and learning support in the
learning characteristics area, and Korean language, belonging, support in the home,
understanding of father or mother culture, and social five specific areas, including
relations, were investigated.
It is believed that underachieving students of multicultural families do not receive
support from teachers or friends compared to other student groups. Even if the
students of the same multicultural family are in sync with learning in terms of learning
characteristics, there is a learning strategy, and learning support is provided, there is
a possibility that they will not fall into sparse learning.
In the area related to the sense of belonging, the group with children of poor
learning of the general family had a significantly higher average score than the
upper-middle group of children of multicultural families. Also, it was found that
the average score was statistically significantly higher in the upper-middle group of
children of multicultural families compared to the underachieving group of children
of multicultural families. Through this, it was found that children of multicultural
families have a relatively lower sense of belonging than children of general families,
regardless of whether they are sparse in learning or not. In the end, it can be seen that
a sense of belonging acts as an essential factor in improving academic achievement
and forming a sense of identity for children of multicultural families. Therefore,
it was analyzed that it is necessary to provide various institutional supports and
opportunities to increase the sense of belonging to children with sparse learning in
multicultural families.
Table 2 shows the sparse learning factors for children of multicultural families
from the perspective of teachers who have experience teaching children of multicul-
tural families. Lack of communication with parents or lack of interest can be seen
as the most significant factor, followed by lack of Korean proficiency, low motiva-
tion for achievement, lack of preemptive learning, and emotional and psychological

Table 2 Factors of sparse learning among children of multicultural families

Division Ratio (%)
Lack of Korean language skills 20.5
Lack of background knowledge about Korean culture 9.4
Prerequisite learning deficit 10.8
Low intellectual ability 3.1
The low motivation for achievement 11.7
Emotional and psychological difficulties 10.3
Difficulty in forming social relationships such as peer relationships 8.0
Lack of communication with parents, lack of interest and care 22.3
Etc 4.0
198 J. Choi and D. Kim

instability. Since these factors of sparse learning are related to each other, if one
problem is solved, several other problems can be solved. Also, since many factors
occur in schools, it can be judged that the existence and role of bilingual teachers in
schools are vital.

2.2 NEIS (National Education Information System)

NEIS is a nationwide education administration information system used by students,

parents, teachers, and education-related public officials. It connects more than
10,000 elementary, junior high, high school, and special schools nationwide and 195
education-related administrative agencies and the Ministry of Education through the
Internet [6].
The NEIS has been used since 2003, and data such as grades, curriculum activity,
and non-curriculum activities in school, and activities outside of school-related to all
students enrolled in elementary and secondary schools are accumulated. The system
supports 92 school affairs and academic affairs, provides 56 school administration
services, 102 general administration services, 18 online services for the general
public, and online materials used in approximately 4.2 million college admission
courses. In addition to providing public services by linking NEIS with the information
systems of related organizations such as the Ministry of the Interior and Safety and
the Supreme Court, it also establishes and implements education-related policies
using NEIS data.

2.3 Big Data Analysis Method and Procedure

To analyze big data, a predictive analysis method for unstructured data such as
unstructured text, video, and audio, which accounts for 95% of the data to be analyzed,
and structured data is required [7]. The process of extracting insights from big data
consists of two processes: Data Management and Analytics. The data management
process includes Acquisition and Recording, Extraction, Cleaning and Annotation,
Integration, Aggregation and Representation, and Analytics include Modeling and
Analysis, Interpretation [8]. Another big data analysis method is descriptive analytics,
inquisitive analytics, predictive analytics, prescriptive analytics, and pre-emptive
analytics [2]. Big data analysis methods include descriptive analytics, inquisitive
analytics, predictive analytics, prescriptive analytics, and pre-emptive analytics, as
shown in Fig. 1 [9].
Inquisitive analytics is a survey to verify/reject data such as statistical analysis
and factor analysis. Predictive analytics relates to predictive and statistical modeling
to determine future possibilities. Prescriptive analytics is about optimization and
Development of a Model for Predicting the Demand … 199

Fig. 1 Classification of types of big data analytical methods

randomization testing to evaluate how to improve service levels while reducing costs.
Pre-emptive analytics can take preventive action against events that could have an
undesirable impact on performance.
To analyze big data, the procedures of data collection, data cleaning, data analysis,
and visualization are generally followed. The data collection step may use a method
of using an open application programming interface (Open API), a web crawling, or
a web scraping method. The data cleaning step is a process of purifying the collected
data into a form that is easy to analyze. It is more essential for unstructured data than
structured data, and text mining is mainly used. In the data analysis and visualization
stage, data is analyzed through Opinion Mining, Social Network Analysis, Cluster
Analysis, and Classification, and Word Cloud and pyLDAvis are used to visualize
the analysis results.

3 Analytical Model Development

3.1 Model Development Process

Linear regression, one of the statistical algorithms, is widely used to analyze big data
[10–12]. Big data can be effectively analyzed by applying machine learning or deep
learning, so consider this and develop a data analysis model in the order shown in
Fig. 2.
200 J. Choi and D. Kim

Fig. 2 Classification of types of big data analytical methods

• Design Items—Design optimization items (Rule Set, algorithm, features, etc.)

to match the characteristics of the target model for machine learning-based data
analysis
• Correction—Correction for optimization
• Data set design—Data Set (Model) design through service analysis
• Validation and optimization—Model validation and algorithm optimization
through training and test data sets
• Setting parameters—Setting optimized model parameters
• Optimal model design—Optimal model design through test data.
Data to be used in the analysis model, the number of elementary schools,
the number of children from multicultural families, and the number of bilingual
teachers were obtained from two data sources. One was provided directly by the
Gyeongsangnam-do Office of Education, and the other was obtained from NEIS.
Comparing the two acquired data, it performs a refining operation for those with
inconsistent or overlapping contents and those with different data formats. In order
to evaluate the optimal algorithm by considering that the response variable in the
optimal model is a numeric type, scikit-learn’s SVM (Support Vector Machine)
kernel SVC (Support Vector Classifier) provides the optimal choose an algorithm.
This process is implemented with the code shown in Table 3.

Table 3 Machine learning algorithm evaluation code

Development of a Model for Predicting the Demand … 201

Table 4 Elementary school, students, and bilingual teachers by year

Year Province Number of schools Number of students Number of teachers
2018 Gyeongsangnam-do 156 639 45
2017 165 638 50
2016 157 587 43
2015 147 542 40
2014 174 417 37
2013 109 243 32

3.2 Data Set Design

Children of multicultural families whose mothers or fathers are foreigners may have
lower academic achievement due to their foreign mothers or foreign fathers who
are not familiar with the Korean language, as compared to other students of the
same age. Thus, objective and technical means are needed to establish a reasonable
budgetary basis in predicting the demand for the smooth supply and demand of
bilingual teachers who will teach children of multicultural families.
This study aims to predict the number of bilingual teachers using the number
of schools and students based on cumulative data on bilingual teachers from 2013
to 2018, taking elementary schools in Gyeongsangnam-do as a sample. Table 4
summarizes the data received through the Gyeongsangnam-do Office of Education
and the data extracted from the NEIS.

3.3 Analysis Model Implementation

Since it is a model for predicting the number of bilingual teachers, linear regression
analysis is the best explanation or prediction of the numeric dependent variable, so
the data analysis model uses a linear regression analysis algorithm.
Statistical results of the final analysis result are generated using ‘statsmodels’ so
that statistical analysis and time series analysis previously performed in R can be
performed in the Python programming language.
‘Statsmodels’ is a Python module that provides classes and functions for the
estimation of many different statistical models, as well as for conducting statistical
tests, and statistical data exploration. Pandas is used to effectively process data in
a two-dimensional matrix structure, such as the number of bilingual teachers for
the number of children in a multicultural family, or the number of teachers in a
bilingual language for the number of schools in the jurisdiction. For the statistical
model, Ordinary Least Squares (OLS) was used, and R-Squared was used for model
verification.
A part of the source code of the data analysis model is shown in Table 5.
202 J. Choi and D. Kim

Table 5 Data analysis model code

4 Analysis Model Application Result

After training an analysis model using related data sets such as the number of schools
belonging to the Gyeongsangnam-do Office of Education from 2013 to 2018, the
number of students, and the number of bilingual teachers, the result of applying the
model by inputting data from the Gyeongsangnam-do Office of Education in 2015
Development of a Model for Predicting the Demand … 203

is R for model performance measurement. The R-squared value is 0.874, and the
adjusted R-squared value shows 0.79. As the R-squared value represents a value
close to 1, it can be seen that the predicted value of the number of bilingual teachers
who performed linear regression analysis is converging within the expected range
through data analysis.
The intercept value is 21.9439, β0 = 0.0383, β1 = −0.0022, and the predicted
value when inputting the 2015 data of 542 students and 147 schools are 42.36321631,
which is the actual number of bilingual teachers in 2015, 40.
There was a slight difference between the number of bilingual teachers and the
actual number of teachers by prediction, but this is because the size of the data set
used for learning is not large enough. Therefore, the prediction accuracy can be
improved by reinforcing the training data set.
Tables 6 is the results of applying the data analysis model. From this, it can be
seen that the big data analysis model presented in this study is significant.

5 Conclusion

As the use of big data and artificial intelligence technologies is becoming more active
in various fields of society, education is also attempting to apply these technologies
in various forms.
Due to internationalization, the number of multicultural families in every country
is on the rise. Children of these multicultural families were found to have relatively
lower academic achievement through education than those of typical families due to
factors such as language, culture, sense of belonging, religion, and environment. To
solve this problem, the need for bilingual teachers is very high. However, due to the
lack of awareness of the importance of bilingual teachers in each region, it is difficult
to secure a budget for hiring them.
To solve such a real problem, a linear regression analysis algorithm was applied,
a data analysis model was developed in Python programming language using pandas
and scikit-learn libraries, and the model was verified with R-Squared. The data anal-
ysis model predicted the appropriate number of bilingual teachers by entering the
number of multicultural family students and the number of bilingual teachers in
elementary schools belonging to the Gyeongsangnam-do Office of Education in
2015. It can be determined that the data analysis model is appropriate by deriving
42.36321631 people.
In this way, when the predicted value is derived based on the data, the need for
the number of bilingual teachers can be raised to the budget department based on the
objective predicted value. Through this, it is expected that the appropriate number of
bilingual teachers to support children of multicultural families who have difficulties
in learning and adapting to society will be maintained.
NEIS accumulates a vast amount of data, but there are not many cases of applying
artificial intelligence technologies such as big data and machine learning to policy-
making in administrative agencies using these data and solving various problems
204 J. Choi and D. Kim

Table 6 Results of bilingual teacher count prediction mode

arising in the educational field. Therefore, this study can be seen as meaningful as
presenting an example of using the vast amount of data held by the government in
the policy. The model of this study will be further developed so that in the future, it
can be used by teachers, parents, and students in the process of college admission.
Development of a Model for Predicting the Demand … 205

References

1. Oxford advanced learner’s dictionary. www.oxfordlearnersdictionaries.com/definition/english/

education?q=education
2. Oh SC, Chang KS, Goo YS (2013) A study on educational support for low-performing students
with multicultural backgrounds: Research Report RRI 2013-2. Korea Institute of Curriculum
and Evaluation
3. Banks JA, McGee Banks CA (2019) Multicultural education: issues and perspectives, 10th
edn. Wiley
4. Karacabey MF, Ozdere M, Bozkus K (2019) The attitudes of teachers towards multicultural
education. Eur J Edu Res 8(1):383–393
5. Ministry of gender equality and family: statistics on multicultural families
6. Ministry of education: introduction to NEIS. www.neis.go.kr/pas_mms_nv88_001.do
7. Gandomi A, Haider M (2015) Beyond the hype: Big data concepts, methods, and analytics. Int
J Inform Manag 35(2):137–144
8. Labrinidis A, Jagadish HV (2012) Challenges and opportunities with big data. Proc VLDB
Endowment 5(12):2032–2033
9. Sivarajah U, Kamal MM, Irani Z, Weerakkody V (2017) Critical analysis of Big Data challenges
and analytical methods. J Bus Res 70:263–286
10. Genuer R, Poggi JM, Malot CT, Vialaneix NV (2017) Random forests for big data. Big Data
Res 9:28–46
11. Saber AY, Alam AKMR (2017) Short term load forecasting using multiple linear regression
for big data. In: 2017 IEEE symposium series on computational intelligence (SSCI), pp 1–6
12. Wang H, Yang M, Stufken J (2019) Information-based optimal subdata selection for big data
linear regression. J Am Stat Assoc 114(525):393–405
Research on Implementation of User
Authentication Based on Gesture
Recognition of Human

Jungseon Oh, Joongyoung Choi, Kwansik Moon, and Kyoungho Lee

Abstract Biometrics technology has taken the authentication technology one step
further, and the solutions, based mainly on static characteristics (fingerprint, face,
retina, etc.), have been commercialized. But there is a possibility that this technology
also can be maliciously bypassed. In that respect, research on new authentication
technology based on dynamic characteristics (gesture, walking, keystrokes, etc.) are
being underway. In the national security facilities such as nuclear power plants and
power generation that require high security level, they are actively adopting biometric
authentication technology based on FIDO certification standards As security breach
incidents become more sophisticated and the scale of damage increases, stronger
authentication methods are required. To achieve this goal, it is expected that an
authentication scheme that combines biometric techniques and existing authentica-
tion methods be studied further. In this paper, we propose a method to implement
such biometric authentication based on gestures (dynamic characteristics) and body
characteristics (static characteristics) and suggest ways to use them.

Keywords Biometrics · Gesture recognition · User authentication · Facial

recognition

J. Oh
KEPCO, ICT Planning, Seoul, Republic of Korea
e-mail: [email protected]
J. Choi
Defense Information System Management Group, KIDA, Seoul, Republic of Korea
e-mail: [email protected]
K. Moon
National Assembly, Seoul, Republic of Korea
e-mail: [email protected]
K. Lee (B)
Graduate School of Information Security, Korea University, Seoul, Republic of Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 207
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_16
208 J. Oh et al.

1 Introduction

As a conventional authentication method, recognizing a user by remembering a pass-

word or possessing an OTP generator (apparatus) has been widely used. Recently,
biometrics (biometrics) has been introduced to strengthen the authentication means.
Smartphone users can use PIN, pattern, fingerprint recognition, and face recognition
(some devices) as a basic authentication method in a device security manner. Since
personal information (photos, contact information, authentication information, and
the like) are mainly stored in a smartphone device, an enhanced security authenti-
cation means should be provided. For personal information or data with restricted
access, it is necessary to confirm that the user is an authorized user. In addition
to password authentication, it is common to verify identity with OTP or biometric
authentication as a secondary authentication.
In the case of PIN and OTP authentication methods, as a second-step authentica-
tion technology, the device must always be possessed (owner-based authentication),
whereas in biometric authentication, the user himself becomes the authentication
means and no other device is required. Also, while authentication information such
as a password may be able to be hacked through several ways, biometric information
is always changed each time at the time of authentication and has a characteristic
that it is difficult to copy. In that respect, biometrics is recognized as an effective
security authentication method. Recently the number of cyber-attacks on power insti-
tutes such as KHNP and KEPCO has reached to 1000 over five years. If even one
attack succeeds, it can cause enormous damage such as paralyzing the power supply
system. In this regard, ways to strengthen the access control are being studied, and
the necessity of biometric authentication as a means of multi-factor authentication
is highlighted [1].
This article introduces general authentication factors and biometric authentica-
tion, and overviews user authentication based on gestures and body characteris-
tics as a dynamic biometric authentication method and attempts to describe the
implementation scheme.

2 Background

In the past, personal and business relationships were formed based on human trust
(user authentication era), but now, machines automatically process trust relationships
instead (service authentication era). The first step in building these trusts is to handle
right identification, which is called authentication. This section overviews authen-
tication factors and technologies as a background theory and describes biometrics.
We will also consider gesture recognition as a dynamic biometric method.
Research on Implementation of User Authentication Based … 209

Table 1 The types of multi-factor authentications

Factors Criteria Characteristics Examples
Knowledge factor What you know No device needed password, virtual keyboard
Easy to use image pattern
Risk of forgetting
Ownership factor What you have Device required IC card, OTP
Risk of loss
High security
Inherent factor What you are Recognition rate problem Biometrics
Authenticated by human
presence
Maintenance cost incurred
High security

2.1 Authentication Factors

As a necessary condition for using an internet service, authentication is a technology

for proving that the user is a legitimate user. It conceptually comes with three tech-
nical elements. The first element uses what the user knows as an authentication
factor, the second uses the user’s possession as an authentication factor, and the
third uses the user’s physical and behavioral characteristics. It can be seen separately
according to the form used [2, 3]. That is, it can be classified into knowledge-based,
possession-based, and presence(biological)-based authentication. And the authen-
tication techniques can be implemented by combining three authentication factors
with each other (Table 1).

2.2 Single-/Multi-Factor Authentication

Power control centers and national security research facilities often require higher
level of authentication technology. This is because the ripple effect of damage is
tremendous when a security breach occurs.
In the field such as financial services, they often require higher level of authenti-
cation technology. This is because security breaches are the most frequent and the
ramifications of damage are high.
In these systems, user authentication may be implemented in various ways, but the
most classic and frequently used method is ID and password authentication. In this
method, the password is registered by the user first, and is encrypted in the service
server using a hash algorithm and stored in the DB. When the service request is
made, the user is confirmed to be a valid user by the password. Thus, authenticating
only a single factor is referred to as single-factor authentication.
In contrast, multi-factor authentication uses a combination of two or more authen-
tication factor technologies. For example, an additional popular OTP authentication
210 J. Oh et al.

device can be used to prove that the user is a legitimate user, who needs to go through a
two-step process of passing OTP authentication again, after the successful password
authentication [4].

2.3 Biometrics

2.3.1 Definition

Biometrics technology is to identify and authenticate an individual by extracting a

user’s physical and behavioral characteristics using a sensor device [5].
For this reason, the biometric authentication element must be unique to every
person and must have a feature that can be used to quantify the extracted data using
the sensor without any change.
For example, fingerprint and iris recognition correspond to this feature. Everyone
has a fingerprint, but each is unique, making it easier to identify an individual and
using a digital device to easily extract body features. Even before the use of digital
sensors, it has been used to identify criminals using fingerprints in a classic way in
criminal investigation.

2.3.2 Biometric Authentication Technology Classification

Biometric authentication technology can be distinguished by using physical and

behavioral features [6, 7] (Fig. 1).

Physical Characteristics

In this method, authentication is performed using physical characteristics that are

expressed by an individual body.
Fingerprints acquire a digital image using a dedicated sensor and recognize the
user. They are classified into optical, capacitive, and ultrasonic fingerprints by the
technology.
In the case of face recognition, it is a bio-recognition technique with the least
resistance by human. And it is implemented by analyzing and recognizing each part
having a characteristic face contour such as a nose, a mouth, eyebrows, and a chin
in a mounted face.
In the case of iris recognition, it authenticates a user using the iris, which locates
between the central black pupil and the white of the eye.
In the case of vein recognition, it uses the fact that is proven to show a different
pattern for each person by using an after-glow by seeing a blood vessel with infrared
rays and not being visible to the naked eye.
Research on Implementation of User Authentication Based … 211

Fig. 1 Biometric authentication scheme can be classified by physiological and behavioral

characteristics

Behavioral Characteristics

This method performs authentication using a specific pattern that is distinguished by

the habit for everyone. There is voice recognition (speaker detection), walking (gait)
recognition, and signature recognition, etc.
Speech recognition is a method of identifying a person based on phonetic char-
acteristics, by extracting and analyzing features from a waveform of speech data
extracted from a human voice.
Gait recognition is a method, in which a human footprint acts like a fingerprint,
using a personal characteristic of a gait. This technique is being studied recently.
Besides, there is a method of extracting the handwriting pressure, the tilt of the
pen, and the method of extracting the characteristics of the character shape.
Also, other research measures the brain waves of a person to identify the area
activated when he is thinking, and this is used as an authentication factor.
212 J. Oh et al.

2.4 Biometrics

2.4.1 Background

Gesture recognition-based user authentication is a technology for authenticating a

user by automatically recognizing a person’s face-gesture, hand-gesture, and other
actions. At present, many research experiments have been conducted mainly on hand
and face gestures [8].

2.4.2 User Authentication

Authentication process begins with the collection of information about identifiable

features, which must be preceded by gesture recognition technology. It involves
dynamically recognizing human gestures, which needs a sensor that can do it.
In the past, a method of attaching to the human body was studied, but nowadays,
a depth camera (3D Depth Camera) is widely used, because it can be easily obtained,
and it has advantages of versatility and convenience.
Gesture recognition using a depth camera analyzes data with a set of frames and
tracking points that capture the motion. For example, in the case of Kinect, 30 frames
are captured per second, and the points to be traced representing the movement of
about 20 or more body-regions from the skeleton in the coordinate space (Fig.2).
The tracking point is where we are interested in biometric information, and we
can use DTW (Dynamic Time Warping) [9] or HMM (Hidden Markov Model) as an
algorithm to analyze this [10, 11].
These algorithms extract a unique pattern of behavior per each person, and we
can determine the identity based on the similarity with the pattern at the time of
authentication.

3 Implementation Method of Gesture Recognition-Based

Authentication System

3.1 Overview

When it comes to implementing the biometric-based authentication system described

above, it will be possible to enhance authentication and improve usability by
combining multiple techniques.
The following figure shows a schematic diagram of a user authentication system
that combines gesture recognition and other authentication means.
A representative example of the application of it is to enter a security zone. Gener-
ally, the entrance and exit of a national security facility requires the strongest level
user authentication.
Research on Implementation of User Authentication Based … 213

Fig. 2 Gesture recognition by using depth camera. (from Microsoft Kinect for Windows SDK)

First, a depth camera and screen terminal are installed at the entrance gate so that
user input can be received. Additional authentication is performed based on infor-
mation extracted from the user’s wearable device, which can be added or disabled
as needed.
After the first authentication, a method for improving user convenience by re-
authenticating only the user’s gesture, excluding terminal input, can be considered.
Until it is near the access gate, the wearable device collects and recognizes the
user’s biometric information (gait). The sensors (depth camera) are used for face
recognition and voice recognition. Also, the touch of the screen(screen terminal),
and the gesture of a mouse can be used.
Frequently, entrance and exit of nation security facilities use fingerprint recogni-
tion. As a means of strengthening authentication, it is possible to selectively combine
these elements so that differential access control can be controlled according to the
reinforced security policy.
214 J. Oh et al.

3.2 Implementation of Dynamic, Static Biometrics

As described above, the target of biometric information recognition will be the touch
screen of the terminal, usage pattern of the mouse, voice, facial-gesture, and hand-
gesture.
Recognition of individual gestures and biometric information involves a common
process of collecting information from individual sensors and authenticating through
learning and inference processes.
If a personalized model is well learned through this process, unauthorized access
to national security facilities cannot be bypassed due to the difficulty of imitation.

3.3 Implementation Method of Access Authentication System

3.3.1 Screen Touch Recognition

From Fig. 3, when touching the screen-terminal with a finger or palm, it is possible to
consider a method to recognize the user’s gesture by mapping the position touched
on the screen to a two-dimensional coordinate plane and tracking the movement
(Fig. 4).

Fig. 3 Conceptual diagram of gesture recognition-based authentication system

Research on Implementation of User Authentication Based … 215

Fig. 4 A general mechanism that recognizes user behaviors, collects and learns data, and establishes
an authentication model using biometric sensors

Fig. 5 Screen touch pattern examples

That method can be implemented, most easily, by the same way as the pattern-
lock, used when unlocking the screen of the smart-phone. This is a method in which
several screen touch patterns are defined, and a password is created by combining
them. It is the most versatile and easy to implement. However, the difficulty may
arises depending on the number of combinations (complexity) and the recognition
rate of the touch screen.
Figure 5 shows some examples of the screen touch pattern. Since the degree of
freedom is higher, as an extended authentication method, than the pattern-lock of the
smart-phone, the benefit of enhancing security can be obtained.

3.3.2 Mouse Gesture Recognition

As with the screen touch method, a method for recognizing the unique movement
pattern of the user using the mouse may be constructed in the same manner as the
screen touch described above.
216 J. Oh et al.

3.3.3 Hand Gesture Recognition

This method uses the information obtained by extracting the contours of the user’s
arms and hands from a short distance using the depth camera of Fig. 3, and to extract
characteristic elements and learn a recognition model.
In implementation, two authentication model, using a static posture and using a
sequence of continuous movements, can be considered.
In the former case, when a gesture meaning the same symbol is taken, the user is
authenticated by distinguishing the static captured image.
In the latter case, a similarity between continuous motions is identified through a
learning process that follows a predefined sequence of motions.
As a dynamic element of hand movement, continuous movements such as left
and right movement, rotation, finger up and folding are patterned and recognized.
But in actual use, learning process will be important. Such as displaying pre-defined
gestures on the screen, and the user following them repeatedly until the accuracy
reaches the minimum critical point.

3.3.4 Face Recognition

Face recognition has already been commercialized to a considerable area, with the
rapid development of the field of image recognition.
We can consider two different methods, of statically extracting facial features or
dynamically using facial gestures.
In the case of face recognition, since there is little resistance from user which
requires minimal action from the user, it can be expected that the synergy effect of
co-operation with others.

3.3.5 Gait Recognition

In the gait recognition, a camera or a wearable device can be used. In this paper, the
wearable device will be described.
There are various biometric authentication factors using sensors of wearable
devices, but this paper introduces an authentication method in connection with gait
recognition [12]. When using a camera, it is necessary to install multiple cameras
to shoot, which is difficult to be used for that recognition. Using a wearable device
attached to the user solves such problems. The sensors used for the gait analysis
can be various, and sensors such as an accelerometer and a gyroscope can be used.
With the products released to the market, you can choose from a selection of built-in
sensors that can be used for analysis.
The analysis will analyze the major joint movements, joint angles and forces and
momentum. In addition, an electrocardiogram is measured, and the movement of a
muscle is used, when gait is analyzed. In practice, it is possible to attach the device
to the waist, pocket, or ankle for user authentication.
Research on Implementation of User Authentication Based … 217

4 Application Scenario

Biometric authentication technology is rapidly growing with the embedding on smart

phones and can be utilized in various fields with the expansion of IoT and wearable
devices.
In the power industry, various service projects such as smart grid, smart distri-
bution, remote monitoring of unmanned substation, and intelligent CCTV are
being promoted. In order to perform these services securely and safely, biometric
authentication technology can be introduced.
For example, when an authentication is required to obtain information in a trans-
mission and distribution facility environment wearing insulated gloves, an alterna-
tive means can be utilized by recognizing an engineer’s arm movement, which is
convenient and safe to a person with limited behavior.
In addition, it can be used to provide a differentiated user experience. In the case
of a facility engineer with a lot of outdoor work, gesture and voice authentication
can provide more reliability and satisfaction.

5 Conclusion

In this paper, we described how to implement user authentication based on gestures

and recognition of body characteristics as a biometric authentication.
Biometric authentication is unique to a user, difficult to duplicate, and charac-
terized by non-identical authentication data each time, and is actively being used
technology that can enhance security.
Static biometrics (such as fingerprints) are widely used today by recognizing the
unique characteristics of the human-user but have the problem that it cannot be used
if the fingerprint is duplicated or damaged.
Research on biometric authentication technology based on behavioral character-
istics has been studied. In this paper, we have described how to recognize gestures
for authentication.
In this paper, an authentication system for access control is considered as an
example. The features are extracted from user-specific actions such as touching the
screen, using a mouse, and hand movements, and user walking using wearable devices
and authenticated for that user by pattern analysis.
In addition, it can be considered to construct a multi-level (multi-factor) authenti-
cation system by linking other technologies using face recognition and voice recog-
nition. In actual use, it is possible to consider putting the difference on the level of
combining the authentication according to authority, and to re-authenticate only by
gait or some gestures to provide user convenience.
218 J. Oh et al.

On the other hand, biometrics security standard or guideline will be needed for
the procurement of it. It is expected that discussion will be made on the quantitative
measure of security strength and indicators, and this will be left as a future research
topic.

References

1. https://www.asiae.co.kr/article/2019092608122663584
2. Van Den Broek EL (2010) Beyond biometrics. Procedia Comput Sci 1(1):2511–2519
3. Apelbaum Y (2007) User Authentication principles, theory and practice. Fuji Technology Press
4. F. F. I. E. Council (2011) FFIEC releases supplemental guidance on internet banking
authentication. DC: Federal Deposit Insurance Corp. (FDIC)
5. Mir A, Rubab S, Jhat Z (2011) Biometrics verification: a literature survey. Int J Computi ICT
Res 5(2):67–80
6. Jain AK, Hong L, Pankanti S, Bolle R (1997) An identity-authentication system using
fingerprints. Proc IEEE 85(9):1365–1388
7. Blackburn D, Miles C, Wing B (2009) Biometrics “foundation documents”. Technical report,
DTIC Document
8. Ducray B, Cobourne S, Mayes K, Markantonakis K (2015) Authentication based on a change-
able biometric using gesture recognition with the kinectTM . In: 2015 international conference
on biometrics (ICB), pp 38–45. IEEE
9. Ten Holt G, Reinders M, Hendriks E (2007) Multi-dimensional dynamic time warping for
gesture recognition. In: Thirteenth annual conference of the Advanced School for Computing
and Imaging, vol 300
10. Huang X, Ariki Y, Jack MA (1990) Hidden Markov models for speech recognition. Edinburgh
University Press
11. Rabiner LR (1989) A tutorial on hidden markov models and selected applications in speech
recognition. Proc IEEE 77(2):257–286
12. Winter DA (1991) Biomechanics and motor control of human gait: normal, elderly and
pathological. Waterloo Biomechanics Press: Waterloo, Ontario
AI TTS Smartphone App
for Communication of Speech Impaired
People

Hanyoung Lee and Deawoo Park

Abstract Due to COVID-19 in August 2020, the announcement by the Central

Disaster and Safety Countermeasure Headquarters is being made regular every day.
Sign language interpretation for the hearing impaired is broadcast on TV next to
the speaker, but language communication for each individual for specific quaran-
tine is insufficient. In this paper, AI construct motion recognition of sign language
interpretation as Big Data, extract features of motion recognition data, and apply it
to AI machine learning. Design an app to communicate with the hearing impaired
on a smartphone owned by the hearing impaired. When a hearing impaired person
writes TEXT or signs language on a smartphone, the smartphone recognizes it. Sign
language motion is converted to text through motion recognition. The converted text
is delivered to the general public as text to Speech through AI voice recognition.
When an ordinary person speaks on a smartphone by voice, it is converted into text
and displayed to the hearing impaired. This study will be used as a means of trans-
mitting information to the hearing impaired in the Untact era. Develop a method for
speech-impaired persons with disabilities to communicate with ordinary people who
have not mastered sign language, using 4th industrial technology.

Keywords Artificial intelligence · Mobile · Big data · Motion recognition ·

Machine learning · TTS (Text to Speech) · STT (Speech to Text)

1 Introduction

The announcement by the government’s Central Disaster and Safety Countermea-

sures Headquarters due to the Corona 19 (COVID-19) is being made on a daily
basis through television and Internet media. With advanced fourth industrial revo-
lution technology, wireless communication and internet are developed in areas of
COVID-19 infected people to deliver information. A smartphone alone reveals the

H. Lee · D. Park (B)

Department of Convergence Engineering, Hoseo Graduate School of Venture, Seoul, South Korea
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 219
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Industrial
Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8_17
220 H. Lee and D. Park

movements of the COVID-19 confirmed patient and receives COVID-19 informa-

tion through map search and AI voice search voice commands. At this time, disaster
measures are being delivered on TV through sign language for the deaf next to the
presenter of the Central Disaster and Safety Countermeasures Headquarters. The
obligation to deliver information to deaf people is important enough to be specified
by law.
However, communication is not easy for deaf people in real life. This is because
sign language interpreters must accompany them in order to communicate with non-
disabled people who have not learned sign language. As a result, communication
between deaf and non-disabled people is subject to great restrictions. In a rapidly
changing society, privacy is unprotected and only a limited number of people can
talk to them, so they can be placed in blind spots of information. In addition, AAC
(Augmentative and Alternative Communication), the existing supplementary alter-
native communication, was developed to meet the standards of people with develop-
mental disabilities who use picture cards of simple word combinations. Like Fig. 1,
there is an overwhelming majority over other types of disabilities, but the research
required for deaf people is insufficient. After meeting sign language interpreters and
hearing impaired people across the country, there is a need and demand for optimized
AAC that allows hearing impaired people to easily communicate with non-disabled
people [2]. This requires practical research to facilitate communication between the
deaf and the non-disabled.
In this paper, AI machine learning is applied to hearing-impaired people’s smart-
phones to design APP for communication. When hearing-impaired people use TEXT
on their smartphones or sign language, their smartphones recognize their movements
and find the right words. In other words, sign language movements are converted to
TEXT through motion recognition. Translated TEXT is expressed as sound to non-
disabled people through TTS (Text to Speech). Conversely, when a non-disabled

Fig. 1 Population by disability type in 2019 [1]

AI TTS Smartphone App for Communication of Speech Impaired People 221

person speaks to a smartphone through voice, it converts it to TEXT (STT: Speech

to Text) and shows it to deaf people.
This paper research will embody the methods of communicating with ordinary
people who have not mastered sign language in the Untact era through AI, and will
be used as a means of information communication and smooth communication.

2 Relevant Research

Sign Language Law: The sign language law aims to improve the language rights and
quality of life of hearing-impaired and Korean sign language users by revealing that
Korean sign language is the native language of hearing-impaired people with equal
qualifications to the Korean language and laying the foundation for development [3].
Motion Recognition: Four basic elements of sign language (hand motion, hand
shape, hand position, hand orientation) that communicate with speech-impaired
people are identified. The motion recognition data stored in advance is compared
using the similarity method, and the sign language is recognized [4]. Although lip
motion, which is recognized by converting to 3D values to increase the accuracy of
sign language, is best applied, lip motion cannot be developed into a smartphone
application that is easy to use because it can be accompanied by special equipment
such as two infrared cameras and three infrared LED (Light Emitting Diode)s. TTS
(Text To Speech) & STT (Speech To Text): TTS refers to the conversion of letters,
sentences, numbers, symbols, etc. into auditory information generally spoken by
people, while STT refers to the conversion of voice into letters, numbers, symbols,
etc. TTS technology has the advantage of being able to do both visual-dependent
tasks. In addition, it is the easiest and easiest way to deliver information that can
change from time to time, and it is a very efficient means of providing information
for people with speech and hearing impairments [5].
AAC (Augmentative and Alternative Communication): AAC is a system with four
components: symbols, tools, strategies, and techniques. It is divided into low-tech
AAC [6] and high-tech AAC [7] f:by using this to communicate their opinions to
others. Even the current state-of-the-art AAC uses symbols (picture cards) for the
intellectually disabled, so there is a limit to having conversations for communication
that the deaf want.
222 H. Lee and D. Park

3 Motion Recognition of Deaf People Analysis and AI

Design

3.1 Motion Recognition Analysis on Smartphones

With the launch of new products every few years, smartphones are convenient to
use features that reflect the Fourth Industrial Revolution technology. You can search
the Internet, play games, e-commerce, and financial transactions that can only be
done on existing PCs, and you can call or search information using voice commands
because AI voice recognition is possible.
In particular, IoT (Internet of Things) sensors that are detected in smartphones are
becoming more intuitive to develop UI than before by using motion, direction, and
heart rate check functions. Smartphones have become capable of various functions
beyond the role of talking on the phone. Initially, it only detected simple movements
through the camera, but now it can make big data for AI. Enter the result value into
the repetitive motion of the person entering the camera to make it a meaningful action
Fig. 2. In other words, data can be created by entering the sign language used by the
deaf. Different people’s hand gestures may make it difficult to extract accurate and
meaningful words for each movement, but sign language is already used by people

Fig. 2 Recognize ‘home’ through map learning using Google AI

AI TTS Smartphone App for Communication of Speech Impaired People 223

Fig. 3 Guidance learning using AAC for the deaf

with speech and hearing impairments, so it forms the same pattern. Entering the sign
language data of many people is a way to improve the accuracy of sign language
interpretation, as input may vary depending on the size of a person’s body type or
movement Fig. 3.
You can use Google AI to learn sign language through instructional learning, and
AAC allows you to enter your own sign language directly, so that you can save it
in Big Data in the cloud and translate it accurately based on a number of learning
samples. This is also available to those who wish to learn sign language, which could
ultimately be an opportunity to break down the language barriers of the deaf and the
non-disabled.

3.2 AI Machine Learning and TEXT Design

You can map and learn sign language movements on AI and build big data to
understand the meaning of simple sign language movements and print them out
in sentences. Through a combination of learned sign language words, a sentence
can be completed by connecting the set words and TEXT the sentence. Figure 4 is
224 H. Lee and D. Park

Fig. 4 Two different ways

of exemplar extraction

a comparison of the two representative extraction methods. The left side is how to
select one single representative for each motion, and the right side is how to select
multiple representatives.
Even if one representative is selected in the best way from a number of instructional
data, in the end, from a different perspective, only one training data is used for each
motion. To reduce the error of sign language, hydration by various deaf people is
essential, and as more data builds up, more accurate words and sentences can be
generated.
Complete the information collected by Sequence to Sequence, an Encoder-
Decoder model using the RNN (Recurrent Neural Network).
In the case of proper nouns, consonants and vowels are recognized individually
and the sign language as symbols is patterned as words, so it is only a list of mean-
ingless words when data is first entered. In addition, due to the characteristics of
sign language, investigations such as ‘ya’, ‘ah’ are omitted, so the consonants and
vowels are combined in order of words and the corresponding surveys are attached
to complete the string Fig. 5.
The encoder receives the data and compresses the information into a single vector.
The vector at this point is called a context vector, and decoder uses this context vector
to complete the sentence shown above. The disappointing point is that the longer the
data words entered are listed, the less complete the combination is.

3.3 Designed with AI Machine Learning TEXT and TTS

Sentences produced by motion recognition will be printed in Text to Speech (TTS).

Once a character identified by the image is entered, the process of processing the
string is carried out. Locate the voice to match the subsequent string in syllables. It
then leads to the process of finding syllable rhymes and creating speech fragments.
Finally, the pieces entered into the image make a word and complete the sentence
with the words, and each of these speech pieces is synthesized in order and then
printed out.
The work has been advanced when delivering voice to TTS, but so far there are
differences in each TTS program. The core of TTS is how accurately it recognizes
text and reads it naturally without any awkwardness.
AI TTS Smartphone App for Communication of Speech Impaired People 225

Fig. 5 Seq 2seq with full-connected layer in the middle

3.4 Designed with AI Machine Learning STT

The ability of non-disabled people to communicate with hearing-impaired people is

limited to visual representation. One way to get out of that limitation is STT (Speech
to Text).
Anyone can use STT easily with the GCP technology provided by Google. You
can run gcloud and Google APIs in a shell window, or you can connect to Google
Compute Engine (GCE) and run APIs.

4 AI App Design for AAC for the Deaf

4.1 Multi-AAC Development

AAC is a necessary skill for many types of disabled people who have difficulty
communicating with their language. However, the high-tech AAC, which was devel-
oped as a symbol card that allows simple information to be exchanged with simple
clicks, is less suitable for people with hearing impairments. Therefore, research and
development of AAC devices suitable for hearing-impaired people who can use AI
together with symbol cards for simple communication is carried out.
226 H. Lee and D. Park

Fig. 6 Google AI sign language recognition via camera

4.2 Sign Language Recognition Analysis APP

on Smartphones

The AAC, developed as a symbol card for deaf people to have a unique or deeper
conversation, has a distinctive limitation. Google AI allows you to expand the scope
of AAC, which is limited to recognizing and learning sign language Fig. 6.

4.3 AI Machine Learning String Learning Analysis APP

Converts to a string through a learned motion with an input image. Using RNN,
complete the information collected by Seq 2seq with sentences and complete with
TEXT. When the string is complete, it is delivered to non-disabled people by TTS
Fig. 7.

4.4 AI Machine Learning Voice Learning Analysis & TEXT

Display APP

In the process of recognizing voice and converting to TEXT, words are often omitted
or distorted. This is due to the mechanical properties (micro sensitivity, ambient
noise) that require continued development of hardware. STT has a difference in
performance between a directional microphone and a microphone built into a smart-
phone. AI can improve accuracy by carrying out a number of studies through learning
Fig. 8.
AI TTS Smartphone App for Communication of Speech Impaired People 227

Fig. 7 Forward image data

learned by AI to TTS

4.5 Comparison of Information Transfer for the Deaf

See Table 1.

5 Conclusion

To cope with social risk situations such as COVID-19 and disaster accidents, studies
on communication with deaf people are needed. Although existing technologies such
as TTS, STT, and motion recognition exist separately, they are not optimized for the
role of AAC for the deaf.
The research in this paper aims to develop auxiliary devices that can communi-
cate with non-disabled and hearing-impaired people who cannot sign at all. In this
paper, the TTS STT technology is applied to the hearing impaired through the smart-
phone, combining IoT precedence and AI voice recognition learning technologies in
228 H. Lee and D. Park

Fig. 8 Change the voice

learned by AI to TEXT

smartphones using the fourth industrial revolution technology. Through AI machine

learning, a smartphone app was designed to learn the behavior recognition of sign
language of the deaf and to accurately judge, and to design and suggest ways to
use smartphones to communicate with non-disabled people. The study in this paper
could protect the privacy of the deaf and have a positive effect on communication
with the non-disabled due to the development of this device.
For future research, it is necessary to study ways to directly communicate normal
people’s language delivery from smartphones through hearing-impaired people’s
smartphones and directly using IoT hearing aids or user sensors.
AI TTS Smartphone App for Communication of Speech Impaired People 229

Table 1 Comparison of AAC development for the deaf

Existing AI Effect
AAC By symbolic drawing Communicable with sign Extended communication
alone limited in use language and symbols between the deaf and the
non-disabled
Motion Sign language Machine learning to Basic DB allows disabled
recognition is inaccurate recognize behavior and people to teach themselves
and serviceability is make more accurate and provides learning
limited sentences opportunities for those who
want to learn sign language.
TTS In the context of Natural pronunciation Natural conversation possible
pronouncing the exact
words resulting in
heterogeneity
STT High missing words and Machine learning reduces Conveniently communicate
error rates word missing and with non-disabled people
typographical errors

References

1. datasom. http://www.datasom.co.kr/news/articleView.html?idxno=105467 (2020)

2. Hearing impaired; Gapyeong (Kim and 12 others) Sign language interpreter: (Lim and 12
others) (2020)
3. Korea Legal Information Center; http://www.law.go.kr/law/Korean Sign Language Method
(2020)
4. Lee B-H, Lee K-S (2004) 2004 Recognition of finger language using image from PC camera.
102–104
5. Kim T, Kim B, Choi D, Lee Y (2012) Implementation of Korean TTS service on Android OS.
9–16 (2012)
6. Special Education Blog; https://m.blog.naver.com/bjs718/221797984500 (2020)
7. https://www.pacific.edu/about-pacific/newsroom/2018/october-2018/aac-camp-creates-
space-for-communication.html (2018)
8. Heo D, Kim C, Kim S, Song B (2019) Comparative study on the speech recognition assistant
device for participation in education of hearing impaired. 101–107
9. Kwon S, Choi Y, Lim S, Joung S (2016) A implementation of user exercise motion recognition
system using smart-phone. 396–402
10. Yun T-J, Seo H-J, Kim D-H (2017) Text/voice recognition & translation application develop-
ment using open-source. 426–427
Author Index

A L
Ahn, Sangho, 179 Lee, Hanyoung, 219
Lee, Kyoungho, 207
Lee, Seulki, 39
B Lee, Siyoung, 95
Byun, Harim, 1 Lim, Donghyun, 165
Lim, Euntaek, 81

C M
Choi, Jinmyung, 195 Moon, Jungkyu, 109
Choi, Joongyoung, 207 Moon, Kwansik, 207
Cho, SangHyun, 53, 67

O
G Oh, Jungseon, 207
Gim, GwangYong, 53, 67, 96 Oh, Saeyong, 53, 67
Gim, Simon, 81 Oh, Youngjoo, 151

H P
Han, SungHwa, 53 Park, Deawoo, 109, 123, 137, 151, 165, 219
Han, Suyeon, 137 Park, Jongwoo, 1, 13

R
J Rou, Hogun, 67
Jo, Donghyuk, 25, 39

S
K Seo, Sungwoong, 95
Kang, Juyoung, 179
Kim, Dooyeon, 195
Kim, Yongmuk, 13 V
Ko, Hansol, 179 Van, Hung-Trong, 81
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 231
J. Kim and R. Lee (eds.), Data Science and Digital Transformation in the Fourth Indus-
trial Revolution, Studies in Computational Intelligence 929,
https://doi.org/10.1007/978-3-030-64769-8
232 Author Index

Vo, Thi-Thanh-Thao, 81 Y
Yang, Jinyoung, 25
W Yoo, Hyunjae, 95
Wang, Hyoboon, 25 Yoo, Hyunju, 123