IS2184 Information System Management --Coursework March 2023

Application of Machine Learning for Enterprise Research

BY ZHANG ZHONGSHI

Abstract
Handling and managing vast amounts of data is vital and complex for all
organisations. Managers use decision support systems and big data analytics to
support and improve organisational performance by analyzing large amounts of data
and providing insights that inform strategic and operational decisions. The objective
of this essay is to examine various ways Machine Learning (ML) could promote
organisation's performance in this data-flourishing era. This essay also examines the
challenges that organisation may face and solutions for issues.

1. Introduction
Decision support systems bring together data and knowledge from different
fields and sources to provide with information beyond the routine reports and
summaries. In doing so, they aim to help people make informed decisions. Big data
analytics could help organizations to discover valuable insights from their data,
leading to the discovery of new trends and information.

Machine Learning is a big data analytic tool that uses various algorithms to
analyze large or small of data sets. In this essay, we have chosen machine learning as
a data analytic technique to examine methods that could improve organisation’s
performance in the following sections. Organizations could use to make decisions that
will enhance business-related outcomes. However, several organisations have not
successfully deployed ML into their contexts as some data and privacy issue brings
great difficulties to their systems.

2. Context
The paper examines discussion in three areas: Prediction, Computer Vision,
and Customer Relationship Management.
2.1 Prediction
Prediction could help organisation in decision-making. This chapter discusses
prediction in two ways: Exchange Rate Prediction and Stock Price Prediction.
Prediction mostly uses supervised learning. Supervised learning refers to the task of
estimating an output, or prediction value for future events, based on a set of input
variables. The algorithm learns from using previous input data and the respective
output to predict output when given new input data. This technique is used to predict a
continuous numerical value based on prior data, such as exchange rate, stock price
and purchase habit.

2.1.1 Exchange rate

Currency exchange rate forecast can help brokers and organisation make
informed decisions to help to maximise trading returns. For example, financial
firms and businesses that import and export goods which exchange currency
frequently. Exchange rate prediction usually uses Random Forest algorithm as a ML
method. While a Random Forest Algorithm belongs to supervised learning method. It
attempts to forecast the dynamic evolutions of exchange rate in the test data period
after training by using the train data. Table 1 shows a well predicted trend with R 2 of
0.975. Specifically, a high R-squared (0.95 and above), indicates a very good fit. This
proves that ML is able to accurately predict future outcomes for exchange rate. This
could help in decision-making for organisations.
 Table 1:Forecasted and actual trending tested

2.1.2 Stock Price

Stock Price Prediction helps to discover the future trend of company stock
traded in secondary exchange market. Scholars such as Hao (2017) and Chen (2018)
state that both demonstrated the feasibility of ML methods for stock price prediction.
By using GBDTt and FFM, ML can complete the screening of factors that can
significantly affect the stock price. While we could use the idea of composite learning
algorithm of these two models to build a stock factor screening model, and can
effectively screen the factors affecting the stock price, Finally, we could observe that
RF model achieved 0.983 in Table2. This shows strong correlation to the actual results.
Subsequently, brokers could gain significant returns for investment firms through
trading in days.

Table 2:Accuracy
 However, machine learning prediction could not consider as dominant role
for trading. One of the real examples of the failure of machine learning of prediction
in the stock market is the investment firm Citadel Securities. The company uses
advanced trading strategies and techniques to provide liquidity to these markets, by
continuously quoting bid and ask prices for a wide range of securities and taking on
the risk of holding these securities in its own inventory. But Citadel faced criticism for
its use of machine learning algorithms for stock market predictions in 2020, as the
algorithms strongly struggled to predict the real prices and adapt to the volatile and
aggressive market conditions caused by the COVID-19 pandemic.

2.2 Computer vision

Computer vision enables computers to analyse, acquire and understand digital
photographs and images. Moreover, it could give recommendation based on the vast
quantities of images to learn the differences between them.

2.2.1 Disease Diagnosis

Computer-aided detection could assist medical institutions in interpretation of
medical images. ML could improve efficiency and accuracy of medical organisation
as computer-aided detection provide opportunity for an improved decision-making
process. From table3, naive bayes could accurately dianose Diabetes by providing the
accuracy of 95% for classification.

Table 3:ML accuracy for detection of Diabetes disease

 From Table 4, it is observed that RS theory makes 100% accuracy for dengue
disease detection. The result proves that ML helps medical institutions in the effective
diagnosis of diseases for patients.

Table 4:ML accuracy for dengue disease diagnosis.

2.2.2 Classification
Classification refers to a predictive modeling problem where a class label is
predicted for a given example of input data. Managers could train a machine learning
model to classify images based on their content and into the correct categories. For
example, be trained by labeling data of shorts and long pants, and applied it to a data
set containing images of shorts and long pants. In some e-commerce companies such
as Taobao, computers could recognise and classify the product type, while even also
distributing to customers who have interest or searching for them. The obtained
models allowed reaching an accuracy rate of 96% for discriminating between a photo,
a text and a mixed document. Subsequently, image classification enhances the
software experience for customers and enhances the precision of distribution
advertisements for organisations.
2.3 Customer Relationship Management (CRM)
It is beneficial for organisations to predict churn rates and help in designing
effective plans for customer retention. This is to prevent customer switch to another
provider, which lose the market share and potential customers which might provide
sales revenue for organisations.

To predict churn risk, enabling you to proactively address customer

dissatisfaction before it results in a loss of revenue. For example, predicting churn
triggers automatic emails. Organisations could predict the customers who will tend to
move to another provider previously by analyzing their behaviour and using machine
learning to send automatic emails. They can retain them by providing coupons and
their preferred services according to their historical records.

Information systems management issue

While most other studies have only focused on applications and algorithms of
data analytic tools, we have also studied data privacy and data quality issues and their
solutions. These problems influence organisation in their business development and
efficiencies. We have also discussed policies and procedures which could solve issues
or minimize the effect if data breaches. This can address information systems
management issues when or if they occur.
3.1 Privacy problem
Privacy is becoming a big concern in this ML-based era. It also becomes more
difficult for both common users and managers to understand the risk of privacy, select
correct schemes and manage privacy. A real-life example of privacy problem is the
use of facial recognition technology which is a specific application of computer
vision(2.2 Computer Vision). In this scenario, agencies collect and store large
amounts of facial images. However, this use of facial recognition technology raises
concerns about privacy and civil liberties, as individuals’ facial images could be
collected and used without informed consent.

3.2 ML-aided Privacy Protection

There are a lot of privacy protection solutions have been introduced. Recently,
research on machine learning-assisted privacy protection is gaining momentum these
days. For example, the use of Graph neural network (GNN) to generate synthetic
datasets has opened up a new direction for privacy protection research, especially for
unstructured data such as images and videos.

3.3 Data quality and Bias

Managers must ensure that the data used is accurate, complete, and unbiased.
This issue occurs if the data collection technique from populations are inappropriate,
this could lead to model bias. For example, in the case of a model in the banking
industry, if the training data is disproportionately made up of applicants from a certain
race or gender, the model may be more likely to approve or deny loans based on those
characteristics, rather than on the applicant's creditworthiness. Managers must ensure
that the models are able to comply with regulations and that the predictions made by
the models are explainable to regulators and other stakeholders.

3.4 Solution for Weak Data quality and Bias

Managers can take a proactive approach to managing the data used to train
the models, such as using data quality tools to check for errors and using techniques
such as oversampling or undersampling to balance the data, or using techniques such
as fairness constraints, re-sampling, or adversarial training to reduce bias. In addition,
managers should also establish a clear governance and oversight framework to ensure
that the ML models are continuously monitored and improved over time, and that the
results are communicated effectively to all stakeholders. This can include regular
performance evaluations, monitoring data quality and accuracy, and regularly
reviewing and updating the ML models to ensure they are still relevant and accurate.
3.5 Data Breach Response Policies
Data breaches happen in almost all organizations even the most effective
defensive layers. Managers could develop data breach protection policies in order to
restrict access, monitor activity, and respond to threats. Specific pre-prevention could
be implementing multi-factor authentication and data security policies, including
encryption, access control, and data retention. If data breaches happened, we could
contain breaches, assess the damage and notify affected individuals to change
passwords. This could provide support to affected individuals and ensure the breach's
impact was minimized.
Discussion
Cambridge Analytica used sophisticated big data analytics and data mining
tools to influence voter behavior during the 2016 US presidential election. The
organisation obtained personal data from millions of Facebook users without
informed consent. This data was then used to create detailed profiles of individual
voters, which were used to target them with political advertising.

For the privacy issue, Cambridge Analytica used illegal ways to get data and
distribute target political advertising, Facebook fails to protect user data and for
allowing third-party apps to access user information without consent. This is a clear
violation of individuals' privacy rights, as they were not informed about how their
personal data would be used and had no control over how it was used.

Data was not a representative sample of the whole population. This makes the
quality of data biased and inaccurate. The data was obtained through a third-party app
and Facebook, which was mostly used by individuals who were interested in quizzes
and social media platforms. This means that the data obtained by Cambridge
Analytica was not representative of the general population.

Cambridge Analytica scandal highlights the importance of data quality and

privacy issues. Organisation should obtain data through suitable designs, and
individuals' privacy rights should be respected and protected as well.

Conclusion
Although there are many challenges, today Machine Learning is noted as one
of the most evolving industries for data analytic tools. Most companies use ML to
assist with large-grouped data analytics. From medical diagnosis and its developments
to prediction and classification, ML has become an important and highly-used tool.
Through this study, we conclude that ML could help in prediction of stock price,
exchange rate, and customer behaviour. Moreover, it helps to make higher accuracy in
disease diagnosis and image classification through computer vision system which
improves the performance of organisation. However, there are some issues with data
analytic tools such as privacy and data quality issue. Organisation should protect
individuals’ privacy and make sure their collected data is unbiased. This could make
difference between ordinary and outstanding companies.
Critical reflections
This section will discuss ICT governance and Knowledge Management
which are valuable to understanding and reducing the complexity of the essay
previously. A further investigation into ICT governance and Knowledge management
will provide fundamental insight into how organisations should manage data analytic
tool to ensure alignment with organisational strategy. The above-mentioned aspects
would also value add to existing solutions in achieving a successful Machine Learning
system.

Knowledge management is the process of identifying, organising, storing and

disseminating information within an organization. This knowledge could include
technical resources and people skills. While past workers could write down their
collected data and descriptions of skills in Excel and Word to transfer knowledge to
the next employees to help them understand the work and prepare for their role. This
could be improved by machine learning algorithms. One common use of machine
learning in knowledge management is to classify text documents into categories. This
can help you quickly find relevant information and organize your knowledge base.
For example, you can use machine learning algorithms like Naive Bayes or Support
Vector Machines (SVMs) to automatically classify documents based on their content
(2.2.2 Classification).

Furthermore, the use of machine learning to build chatbots and virtual

assistants that could help users find information quickly and easily. It helps managers
to improve decision-making by maintaining and improving access to current and
historical enterprise knowledge. Hence, big data analytics and DSS could help
organisations to maximise efficiency in making the most of their data by enabling the
effective capture, storage, and retrieval of information for knowledge management.

ICT governance could help organisations ensure that the use of big data
analytics aligns with the organization's overall goals and objectives, while also
ensuring the security and privacy of sensitive information. While good ICT
governance will not only ensure this alignment but also ensure necessary economic
and managerial resources for ICT deployments. Moreover, a good ICT governance
system will also ensure continuous maximum output for organisation’s performance
management.

ICT governance is at the centre of managerial interests given increased ethical

concerns. The collapse of Enron in 2001 is a well-known example of the
consequences of poor ICT governance. One of the key issues with Enron's ICT
governance is lack of transparency and accountability. Executives were able to
manipulate the financial data because there were no effective controls in place to
prevent this. In addition, there was a lack of independent oversight of the IT
department, which meant that the company's senior management did not have access
to accurate information about the risks associated with the technology systems. The
collapse of Enron serves as a stark reminder of the importance of effective ICT
governance. Organizations need to have clear policies and procedures in place for
managing their IT resources, and they need to ensure that these policies are adhered to
by all stakeholders. This includes senior management, IT staff, auditors, and external
vendors.

However, there are also some successful examples like Citadel Securities
(2.1.2 Stock price). Citadel Securities include complete ICT governance guidelines of
data management, risk management, compliance, change and infrastructure
management to ensure the integrity and security of trading systems and data. ICT
governance promotes transparency and accountability in the use of technology in
quantitative trading.

ICT governance for Citadel Securities includes policies and procedures related
to the management of changes to technology and processes, such as testing and
validation of new technology and technique for prediction (2.1 Prediction) from
quantitative trading(change management). During the coronavirus pandemic
(2020~2022), ICT governance (change management) makes Citadel Securities double
its profit due to perfect management of changes from ICT governance. Moreover,
their previous structured complete risk and data management system made the trading
system more powerful and competitive. With increasing volatility and trading, Citadel
Securities is also able to align with the organization's objectives, comply with
regulatory requirements, and minimise risks. Hence, ICT governance is needed to
develop and manage plans to guarantee business continuity and hence to provide
effective support to every business activity.
Bibliography

1. Parker, G. (ed.) (1903) Mark anniversary volume. New York: Henry Holt.(Accessed 17 Jan 2023).
2. Zeng, L. L., Xiao, Y. N., & Chen, S. L. (2022). Research on Short-Term Stock Price Trends Based
on ML. iBusiness, 14, 75-94. https://doi.org/10.4236/ib.2022.142006 (Accessed 17 Jan 2023).
3. Tsuji, C. (2022). Exchange Rate Forecasting via a ML Approach. iBusiness, 14, 119-126.
https://doi.org/10.4236/ib.2022.143009 (Accessed 17 Jan 2023).
4. Reinhard Klette (2014). Concise Computer Vision. Springer. ISBN 978-1-4471-6320-6.
5. Journal of Intelligent Learning Systems and Applications, 2011, 3, 90-102
doi:10.4236/jilsa.2011.32011. (Accessed 17 Jan 2023).
6. Fatima, M. and Pasha, M. (2017) Survey of ML Algorithms for Disease Diagnostic. Journal of
Intelligent Learning Systems and Applications, 9, 1-16.https://doi.org/10.4236/jilsa.2017.91001
(Accessed 18 Jan 2023).
7. doi:10.4236/jsea.2012.512b038 , Image Classification using Statistical Learning Methods,
Jassem Mtimet, Hamid Amiri Elisa Costante, Jerry Den Hartog, and Milan Petkovic. 2011. On-line
trust perception: What really matters. In Proceedings of the 1st Workshop on Socio-Technical
Aspects in Security and Trust (STAST’11). 52–59. (Accessed 19 Jan 2023).
8. Elisa Costante, Yuanhao Sun, Milan Petkovic, and Jerry Den Hartog. 2012. A ML solution to
assess privacy policy completeness (short paper). In Proceedings of the ACM Conference on
Computer and CommunicationsSecurity (CCS’12). 91–96.
https://doi.org/10.1145/2381966.2381979 (Accessed 19 Jan 2023).
9. Hosub Lee and Alfred Kobsa. 2017. Privacy preference modeling and prediction in a simulated
campuswide IoT environment. (Accessed 19 Jan 2023).
10. Primal Wijesekera, Arjun Baokar, Ashkan Hosseini, Serge Egelman, David Wagner, and
Konstantin Beznosov. 2015. Android permissions remystified: A field study on contextual integrity.
In Proceedings of the 24th USENIX Security
Symposium (USENIX’15). 499–514. (Accessed 19 Jan 2023).
11. Primal Wijesekera, Arjun Baokar, Lynn Tsai, Joel Reardon, Serge Egelman, David Wagner, and
Konstantin Beznosov. 2017. The Feasibility of Dynamically Granted Permissions: Aligning Mobile
Privacy with User Preferences. In Proceedings of the IEEE Symposium on Security and Privacy
(SP’17). 1077–1093. https://doi.org/10.1109/SP.2017.51 (Accessed 19 Jan 2023).
12. Ebrah, K. and Elnasir, S. (2019) Churn Prediction Using ML and Recommendations Plans for
Telecoms. Journal of Computer and Communications, 7, 33-53. (Accessed 19 Jan 2023).
https://doi.org/10.4236/jcc.2019.711003
13. Journal of Software Engineering and Applications, 2012, 5, 200-203
14. doi:10.4236/jsea.2012.512b038 Psychology, 2014, 5, 2046-2057 Published Online November
2014 in SciRes. http://www.scirp.org/journal/psych (Accessed 19 Jan 2023).
http://dx.doi.org/10.4236/psych.2014.518207
15. International Journal of Network Security & Its Applications (IJNSA), Vol.4, No.1, January 2012
(Accessed 23 Jan 2023).
16. Journal of Information Security, 2021, 12, 34-55 https://www.scirp.org/journal/jis
17. Tsuji, C. (2022). Exchange Rate Forecasting via a ML Approach. iBusiness, 14, 119-126.
https://doi.org/10.4236/ib.2022.143009 (Accessed 23 Jan 2023).
18. https://www.sas.com/en_us/insights/analytics/machine-learning.html#machine-learning-wor
kings (Accessed 23 Jan 2023).
19. https://www.putitforward.com/put-it-forward-predictive-analytics-for-revenue-generation
(Accessed 23 Jan 2023).

Word count
The word count is 2485 excluding titles, labels and bibliography.

-End-