Chapter 1

INTRODUCTION

This chapter outlines and focuses the research field reconnoitered in this doctoral study. In
particular, describing I-Commerce. I-Commerce technique along with recommendation systems
is focusing on optimization techniques with parallel clustering algorithms. The main purposes
along with the organization of this document are articulated.

INTRODUCTION 1

1.1 Background and Introduction to I-Commerce 3

1.2 Types of I-Commerce 5

1.3 Advantages of I-Commerce 7

1.4 Growth of I-Commerce 9

1.5 I-Commerce Tools 11

1.5.1 Recommendation systems 11

1.5.2 Optimization Algorithms 13

1.6 Big Data 17

1.7 Parallel Clustering Algorithm 19

1.8 Challenges in I-Commerce 19

1.9 Current Research Trends In I-Commerce 20

1
The rapid development of Social Networking revolutionary changes our daily lives and global
business, which has been addressed in recent research. Recommendations play a vital role through
which vast amounts of information shared among different internet users every day.
Recommendations can be in the form of reviews, surveys, ratings, spoken words, reference letters
reports from news media, travel guides etc. Amongst large collection of available items
recommender system can predict personalized item for users. The main goal of recommender
system is to select appropriate item based on user interest and requirements. Items can be classified
as videos, books, office products, home appliances mobile accessories etc. The emerging growth
of e-commerce web portals, e.g. Amazon.com, Flipkart.com, Netflix.com, TripAdvisor.com etc.
have created enormous interests in RS 1 research to provide effective recommendations to end-
users. Recommendations are divided into the two approaches namely content-based, where on the
basis of user history and personal data products are recommended to the user. Currently goggle is
using the similar techniques for recommendations. Collaborative filtering, where we take benefit
of the fact that people who had same tastes in the past may also agree on their tastes in the future.
Despite the success of CF2 approach, it suffers from many serious problems related to data sparsity
that affect recommendations’ accuracy.

To enhance the quality of such RS many researchers have implemented different techniques but
different problems are identified. The first one is the long tail problem which concerns items with
only a few ratings. The second is the cold-start problem which has two variants such as new-user
when an unknown user joins the system looking forward recommendations, new-item (called also
first rater problem) in which the RS is unable to deliver to users new items without any rating. The
new-item problem is a special case of the long tail problem. In recent years, research on the
enlargement of clustering techniques has attained notable attention in numerous data mining
applications such as text mining, pattern recognition, web analysis, information retrieval,
visualization and image segmentation. In the user-based recommendation model, the clustering
technique is used to determine a similarity measure to cluster the users based on the similar item-
ratings given by various users. The items systematized based on similarity ratings in a few clusters
of users is recommended to new user who has similar tastes to the user group. Many clustering

1
Recommendation System
2
Collaborative Filtering

2
techniques are used to produce personalized recommendations in the user-based model such as
fuzzy c-means, K-means, SOM etc. With the development of real-time clustering-based
recommender systems, the information dispensation problems make the recommendation process
more complex. The usage of traditional clustering algorithms such as nearest neighbor K-means
clustering algorithm has some drawbacks in obtaining best or optimal solutions for large-scale
application problems. The primary objective is to propose a prediction based parallel Clustering
Algorithm using Optimization Technique for identifying users need and predict the result as per
user’s requirement as well as enhance the relevancy of results.

1.1 Background and Introduction to I-Commerce

In the past two decades, buying and selling of the products are carried over the internet. Online
shopping has changed the lifestyle of the human. Initially in 1994 pizza hut has successfully sell
pizza online and the concept of E-commerce3 is introduced into the picture. Then many researchers
have introduced a new technology for buying the products online. Many researchers also focused
on schemes related to online business to increase the sell over the internet. They have also
introduced new technologies like recommendation systems, online rating etc. Then in 1997, Kevin
Duffey has introduced a concept of M-Commerce4 where buying and selling products take place
through mobile means or handover the E-commerce capabilities into the customer's hand. In 2017
few researchers have introduced a concept of IOT5 Commerce where automated buying of
products takes place without the interference of consumers. For example, if we are using any
consumables like wheat in the home and if it is connected to internet with the help of sensors, if
the wheat level is low automatically sensor will send the message to shopkeeper and he will give
the delivery of product to your home. I-Commerce6 is a combinatory evolution of E-Commerce,
M-Commerce, and IoT-Commerce7, where buying and selling of products take place with the help
of the Internet. In this environment, customers are able to buy or sell the products of their choice
at reasonable prices. As per the researchers, I-Commerce is defined as

3
Electronic Commerce
4
Mobile Commerce
5
Internet of things
6
Internet Commerce
7
Internet of things Commerce

3
 “Providing a software package to business owners which will help for a fast and affordable
way of selling the products and services online with no financial risk by Electronics or Mobile
means”.

It covers auctioning, collaborating with trading partners, making payments, buying and selling of
products, transferring funds, etc. I-Commerce system can help us to conduct research in various
stages with respect to current requirements. As per the architecture is shown in figure 1.1 we can
process large amounts of data using I-Commerce architecture. The slogan said by one of most
famous socialist “By the people, for the people, to the people”. By keeping same concept in mind
and to provide similar facilities to the consumer I-Commerce model is developed. I-commerce
relies on a recommendation system that helps in the growth of the market and increases the profit
of sellers. Architecture of I-Commerce it is divided into 4 parts Consumer, Integration, Analytics
and Data. It is driven by the consumer where Buying and selling of products take place through
the web, mobile, IoT and store channels. For performing this operation various services are used
like Global shopping cart, I-Commerce Front-End, Services API. Then it is integrated into
payment gateways using real-time inventories and product catalogs. Once the process is completed
data is stored into the database, then by using historical data and User reviews, analytical data is
generated. This whole process describes the steps involves in basic architecture of I-Commerce.
The best example to validate I-Commerce architecture stages is trivago.com. Where developer is
parsing all the websites through the analysis phases and gives a most relevant hotel for the
consumers by identifying their tastes. Many other shopping websites are also using a similar
concept to recommend and predict the most relevant products.

4
 Figure 1.1 Architecture of I-Commerce

1.2 Types of I-Commerce

Internet Commerce is categorized into three parts as per the behavior of commerce techniques.

1) E-Commerce
2) M-Commerce
3) IoT Commerce

E-Commerce is a buying and selling of products takes place through electronic means. Modern E-
Commerce uses WWW8 for transactions. E-Commerce transactions include buying of Electronics
gadgets such as mobile phones, Purchasing of Books (Amazon), Purchasing of Music (ITunes
Store). It May employ online shopping of the products through internet or direct to customers via
websites, gathering the data through websites and social media. There are three main areas are
involved in E-Commerce are Electric Markets, Online retailing and Auctions. M-Commerce is a
handing over the E-Commerce capabilities into the customer pocket. Many retailers have started
using M-Commerce like Amazon, Flipkart. It provides a support for enormous services like Mobile
money transfer where money can be transferred online like Google pay app, Mobile ticketing
where we can book the ticket online through mobile like UTS app, Mobile vouchers, coupons and

8
World Wide Web

5
loyalty cards like Phonepay is provide a vouchers or coupons on online shopping or transferring
the money, content purchase or delivery where we can buy ringtones, wallpapers, games through
mobiles, information services like news, stock quotes, sports scores etc. Mobile marketing and
advertisings where online consumers are received promotional messages or vouchers or coupons.

IoT Commerce is Internet of Things commerce where automated buying and selling of the
products takes place via IoT. IoT is communication between any two sensors with the help of
internet. Currently, world is moving with the IoT, where in future buying and selling of the
products will takes place automatically without knowing the user as per the need of users.

Figure 1.2 Wheat Tin IoT Commerce Product

As shown in figure 1.2 IoT Commerce can be used in home automation systems for purchasing
regular households at home. Here, we have connected level sensors to check the level of wheat in
house. If the level of wheat is low in tin, automatically it will send message to buyer and seller.
Seller will wait for the confirmation and if confirmation receives, seller will give the delivery on
the registered address.

As shown in figure 1.3, these types are divided into 4 parts i.e. B2B9, B2C10, C2B11 and C2C12.

9
Business to Business
10
Business to Consumer
11
Consumer to Business
12
Consumer to Consumer

6
 Figure 1.3 Types of I-Commerce
B2B is a web-based economical business model used for selling and buying of products between
two business users. Alibaba or Indiamart is famous examples of B2B category.
B2C is a web-based business model used between business users and consumers, Target.com or
Trivago is a famous example of B2C type.

C2B is a consumer to business model which is a mirror image of B2C. Here, Consumers can sell
a product and buying is to be carried out by business users. Amazon.com is a good example of
C2B type. The last one is a C2C model where buying and selling of products take place between
consumers. OLX.com is a famous example of C2C model. Online buying and selling of products
have a lot of advantages and it creates ease of access for buying or selling of products with
trustworthy users. All this types can be correlated for all types I-Commerce and can be
implemented in real time environment.

1.3 Advantages of I-Commerce

I-Commerce provides vast opportunities for various organizations to promote their business.
Through online business, users can reach to a variety of customers throughout the globe. Reducing
the cost of infrastructure, by availing the various schemes buyer can increase the sales of the

7
product. Organizations can also increase sales by applying various technologies digital marketing
or recommendation techniques.

1.3.1 24 X 7 Buying and Selling Availability

Consumer has a freedom to buy or sell products through online anytime, anywhere that is 24 X 7,
365 days. Customers can purchase a product anytime if customer is not happy with size, color
dimensions or different parameters of products, customer can return the product anytime, as per
the convenience of customer.

1.3.2 Buying the Products with Varieties and Choice of User

I-Commerce is a blessing showered by God to the human being. We can buy all the products
starting from vegetables to furniture, electronics, and gadgets without showing actual presence in
the store. I-Commerce has provided a facility to have all sides or deeper look and feel of products
like Lenskart.com.

1.3.3 Easily Access to Latest Trends

I-Commerce helps us to access the latest trends or new arrivals which are not available in the store,
Because of which we are not able to buy the latest arrivals in the market. I-Commerce has made it
easy to access the latest trends through the internet.

1.3.4 Reliability of Products

Better services are provided while buying products online. There is enough information is
available about products online like price, reviews, detailed descriptions, ratings. It helps to buy
proper products online which leads to the satisfaction of users. Reliability always helps seller to
increase their business. Reliability is to sell the authenticated products to the customer which
satisfies all the requirements off users.

8
1.3.5 Rewards and Offers
There is good service and offers are being provided while buying a product online. Many
commerce sites provide a discount, rewards, offers which helps the customer to buy products at a
cheaper price.

1.3.6 Anytime, Anywhere, Anyone can Become a Seller

C2C model emphasizes on buying or selling of products between consumers. Consumers are able
to sell the products via online interface where infrastructure is to be provided by various web
portals with nominal charges. Olx.com, Quickr.com is a famous example of C2C type.

1.4 Growth of I-Commerce

Most of the customers in today's era are inclined towards the I-Commerce. Business growth
indicates that there are huge growth and rapid development in this area. Recent studies show that
traditional business is almost replaced by I-Commerce. In the next five years, more than 80% of
peoples will choose I-Commerce. Business statistics show that huge growth of I-Commerce is
because of ease of access, availability, variety of products are available and online commerce
awareness about I-Commerce. Several studies and surveys show that growth of online users is
increased in a tremendous manner. Figure 1.4 shows the details of the statistics of internet users
across the globe.

Figure 1.4 Statistics of Internet users according to geographical reason [10]

9
As per the Internet world statistics, approximately 4380 million online users are availing the
facility of Internet. Means 60% of world population is using an internet. Obviously, as the number
of users increases over the globe on internet it is going to dominate traditional business.

As shown in figure 1.5 in 2018 a US I-Commerce company has total sales of $252.69 billion.
Amazon has 49% of sales in the year 2018. The above scenarios focused on growth of online
commerce.

Figure 1.5 Top 10 I-Commerce Companies [12]

10
 Figure 1.6 World wise Retail Sales worldwide [8]

As shown in figure 1.6 it shows the continuous growth of retail sales over the years. As per the
predictions and many researchers till 2021 I-Commerce will reach to the peak of mountains.
Maximum money will be invested, spent and earn from I-Commerce.

1.5 I-Commerce Tools

I-Commerce provides huge opportunities to various industries for promoting their business. In the
world of competition, it is very important to have various techniques to increase the growth of the
business. Researchers have developed many techniques to overcome these issues like
recommendation systems, availing offers and awards on users search, digital media search base
prediction techniques and optimization techniques to extract the relevant products.

1.5.1 Recommendation Systems

Recommendation system is filtering the information on the basis of user behaviors or
recommending the products to user by applying some algorithms or computational techniques. RS
is used in many areas like Online selling, YouTube, Netflix, etc. RS usually extracts the data on
the basis of past behavior of users. It uses either collaborative filtering or content-based filtering.
In a few cases, other RS like Multi-criteria recommendation system or hybrid recommendation
system is also used. These techniques are used to find the interest of users. YouTube is the best
example of RS where on the basis of user historical data YouTube gives new predictions of the
videos to the users. It uses both Collaborative filtering and content-based filtering. Combination
of various RS is called a Hybrid recommendation system.

11
Collaborative filtering is working based on the assumptions that in past users are agreed for buying
the product will agree for the same in future. Previous historical data between peer users and rating
of the products from different users, Product is recommended to users. Figure 1.7 demonstrates
the simple example of collaborative filtering. The process involved in collaborative filtering is
asking a user to rate an item, then system asks user to search the product, at the time of searching
on the basis of few users past behavior it tries to find out the correlation and similarity between
different items and recommend the same product to user.

Figure 1.7 Collaborative filtering

The best example of collaborative filtering is buying a mobile phone on Amazon portal. While
buying a mobile system it suggest us to buy the headphone or mobile back cover on the basis of
previous user history or some users have purchased mobile with respective products. In these
techniques, system uses the nearest neighbor method for finding similar items. Collaborative
filtering is also used in social networking for suggesting friends.

Another common approach of RS is a content-based recommendation system. In content-based RS

is based on product description and user preferences. This method is best if we know the
description of the product but not having an idea of users. It is based on likes and dislikes of users,
it uses a classifier algorithm for recommending products. It also uses the opinion based RS which

12
allows users to enter the views, comments, this data can be used to recommend the products next
time to another user. Popular trends in opinion based RS are sentimental analysis, text mining and
information extracting. The main issue with this type is system is not capable to learn the user
preferences taken from users actions. Here system is recommending the same products which users
have already purchased. To solve this problem many recommender systems are using hybrid RS.

Hybrid RS is combining two or more types of RS together to share the properties. Generally,
Content bases RS and collaborative filtering system is combined together with a few other
approaches. It gives a good result with high accuracy and relevancy. These methods are also used
to solve the common problems of RS like cold start problem and sparsity problem. Netflix is the
best example of hybrid RS. The most important parameters of hybrid RS are combining the score
of different RS numerically, Selecting the recommendations components, features combinations
and augmentation. Recommend the products as per the priority of users is select an appropriate
meta-level to produce finalized output.

1.5.2 Optimization Algorithms

Optimization is a process of enhancing the system to achieve some features and work more
efficiently using less resource. For instance, we can optimize a computer program in terms of space
and time to increase the performance of program. In I-Commerce optimization is used to select the
appropriate product or help the customer for recommending the appropriate product. Various
optimization algorithms are developed to increase the efficiency of I-Commerce. Ant Colony
Optimization, Particle Swarm optimization, Ant bee colony optimization, Grey wolf optimization,
etc. are the techniques used to select the best product from ocean of the Internet.

1.5.2.1 Ant Colony Optimization

It is a probabilistic technique for solving computational problems to find the shortest path. By
cultivating the features of Ants habits into real-time computing for the benefits of humans. In these
techniques, ants are trained to find the food from various resources and the task would be carried
out in different groups. In groups, ants are starting searching from the nest to food sources. If the
food sources are found ants have updated the information to all other ants. While coming back to

13
nest ants are using a pheromone substance to keep a track of food source. Traveling salesman
problem can be solved using ACO 13.
Procedure
ACO_MetaHeuristic
Begin
while (not_termination)
Generate Solutions ()
Daemon Actions ()
Pheromone Update ()
end while
end
In Edge selection, each ant moves from state ‘x’ to state ‘y’. Each ant ‘k’ computes a sets Ak(x) of
feasible solution for each set for each iterations. For ant, k probability is Pxyk from state ‘x’ to state
‘y’. It generally depends on attractiveness ηxy and the trail τ of the move. Trails are updated when
all ants have updated their solutions. The probability of k th ant from state ‘x’ to state ‘y’ is

β
𝑘 (τα
𝑥𝑧 )(η𝑥𝑧 )
𝑝𝑥𝑦 = β ………………….. (1.1)
Σz ϵ allowed y (τα
𝑥𝑧 )(η𝑥𝑧 )

Where τxy is the amount of pheromone deposited from state x to y α and β are the parameter to
control influence to represent the attractiveness and trail level τ xz , ηxz for all the possible
transitions.
To update the pheromone the trails are updated once all the ants have completed the updations are
done as per equation number 1.2.

τxy  (1-ρ) τxy + ∑𝑘 ∆τ𝑘𝑥𝑦 ………………….(1.2)

ρ is the coefficient of pheromone evaporation and Δτkxy is the amount of pheromone deposited by
kth ant. The same concept can be implemented for solving traveling salesman problem.
Steps involved in solving traveling salesman problems are
1. The first rule is to visit each city exactly once.

2. A distant city has few chances of being selected in terms of visibility;

3. If the pheromone trail laid out between any two cities is more, the probability of selection of
edge will be more.

13
Ant Colony Optimization

14
4. Having completed its journey, the ant deposits more pheromones on all edges it traversed, if
the journey is short; If the Journey is completed, More pheromone is deposited on all paths
which are traversed.

5. Trails of pheromones are evaporated after completion of each cycle.

Ant colony optimization can be used in I-Commerce, but the accuracy and relevancy of prediction
of products are less. ACO focuses on quantity of products, not on quality of products. So to solve
the further issues researchers have introduces a new concept called Adaptive ACO. In ACO
searching process is based on positive feedback reinforcement techniques using pheromone
information. But, escape from local optima is more difficult than other meta-heuristics. For solving
the above problem adaptive ACO uses the transition of the shortest path. In this technique, cranky
ants are introduced into the system. Cranky ants select the shortest distance having a lot of
pheromones. The path has lots of pheromones indicates as many-selected path. Using the
recognition techniques achieves control of tradeoff between intensification and diversification.

1.5.2.2 Particle Swarm Optimization Technique

PSO14 is a computational method that optimizes a solution to the problem by performing an

iterative operation to increase quality. It works for population called as a swarm having a candidate
solution (particles). These particles are move along the search-space. The movement of particles
is guided by their own best position in the search space. When an improved position is discovered
it is updated to all the swarms and process is repeated to reach the best solution.

Algorithm for PSO

1) Start
2) for each particle j = 1, ..., S
3) Initialize the particles position with uniformly distributed random vector:
4) xj ~U(blo, bup)
5) Initialize the particle's best known position to its initial position: p j ← xj
6) if f(pj) < f(g) then
7) update the swarm's best known position: g ← pj
8) Initialize the particle's velocity: vj ~ U(-|bup-blo|, |bup-blo|)

14
Particle Swarm Optimization

15
 9) while a termination criterion is not met do:
10) for each particle j = 1, ..., S do
11) for each dimension d = 1, ..., n do Pick random numbers: r p, r0067 ~ U(0,1)
12) Update the particle's velocity: vj,d ← ω vj,d + φp rp (pj,d-xj,d) + φg rg (gd-xj,d)
13) Update the particle's position: xj ← xj + vj
14) if f(xj) < f(pj)
15) then Update the particle's best known position: pi ← xi
16) if f(pi) < f(g) then
17) Update the swarm's best known position: g ← pi
18) Stop

PSO uses two main principal viz. communication and learning. To find the search space and
updated solutions it performs a communication with another agent. While performing
communication function, it focuses on learning to find the stochastic solutions.

1.5.2.3 Grey Wolf Optimization Algorithm

Meta-heuristics techniques are very popular over last 3 decades. Physical activities of animals,
birds are encourages researchers to employ the same concepts in real-life examples. These
techniques are either individual-based or population-based. GWO15 is a technique developed by
Mirjalili (2014) by employing the concept of leadership hierarchy of wolves. The main leader is
called an Alpha (α). Alpha is responsible for decision making. Dominant leader is called as beta
(β). Beta is subordinate wolf help the alphas for decision making. The lower rank wolves are
called as an Omega (ω) and delta (δ). These wolves are responsible for the completion of task
assign by alpha, beta and report to them. The hunting techniques and wolf behavior are
mathematically structured to solve real time engineering problems.

Steps involved in GWO [2]

1) Initialize GWO parameters,

Gs are denoted as a search agent and Gd is denoted as a variable size design. Then

15
Grey Wolf Optimization

16
 Vector A, a, C and Itermax

⃗⃗⃗
A =2. ⃗⃗a Rand 1− a the value of 𝑎
⃗⃗⃗ linearly decreased from 2 to 0

𝐶 = 2. Rand 2

Generate the wolves based on the size of pack randomly

𝑔11 𝑔21 .
Wolves = [𝑔12 . . ]
𝑔𝑠
. . 𝑔𝑔𝑑

Where 𝑔𝑗𝑖 is the initial value of jth pack of ith wolves.

2) Find out the fitness value for each agent

3) Identify the best hunt agent Gα, Gβ, Gδ
4) Renew the location of current hunt using

⃗⃗⃗⃗⃗⃗ + 𝐺2
𝐺1 ⃗⃗⃗⃗⃗⃗ + 𝐺3
⃗⃗⃗⃗⃗⃗
𝑔
⃗⃗⃗ (t+1)= ……………………… (1.3)
3

5) Estimate the fitness value of each hunt

6) Update the value of Gα, Gβ, Gδ
7) Continues the iterations and check for maximum iteration Itermax if yes print the best value.

GWO can be used for many real-time applications like E-Commerce, Fuzzy systems. Most of the
swarm intelligence techniques that are used to solve the optimization problems are not having a
leadership quality. This drawback is overcome by GWO algorithm. Optimization algorithms can
help to find the most optimized dat. But if the size of data is large then we have to use technique
which can give the most optimized result in minimum time with high efficiency. To achieve this
objective the best option is either we can use Hadoop or parallel clustering algorithms.

1.6 Big Data

Term “Big Data” gained momentum in early 2000s. Large amount of structured and unstructured
data that cannot be processed by traditional relational database engines is referred to Big Data.
Big data is complex such that traditional database engines are inadequate to process it. Open-
source framework Like Hadoop is the reason for growth of big data which was created to store and

17
maintain large data sets. Big data involves generating, storing, analyze, retrieving and transferring
large amount of data. The term “Big Data” not only refers to how much data you have but also
what you do with that data. It involves various techniques, frameworks as well as tools. Big data
helps to analyze data, identify patterns and behavior which helps to take decisions in better manner
which results in development of business or organization. Big data is a large amount of data in
which data analyze, visualize, summarize and extract the information using application software.
It is characterized by Volume, Variety, Veracity, Velocity and Value as shown in figure 1.8. In
most of the phases big data works on the concept of Map Reduce. Hadoop is a famous framework
which can be used to process large amount of data. Hadoop is an open source framework that can
be integrated with Java (Eclipse), Python (Anaconda), and R (R Studio).

Characteristics of Big Data

Volume Variety Veracity Velocity Value

Figure 1.8 Characteristics of Big Data

Volume: It is the quantity of generated or stored data. Data can be produced by various ways like
the stock market, airline systems, Natural systems, Transportation, manufacturing, etc. Data size
is transformed from zettabyte to petabyte.

Variety: Data is produced in various formats. It may be structured, Unstructured & semi-structured
data. With traditional systems, it is not possible to manage a large amount of data. Structured data
is a data-oriented in one particular format like tabular formats having rows and columns,
unstructured data is a combination of text, audio, images, videos. And semi-structured data is a
data which is separated by a particular separator like a comma, tab.

Velocity: In today's era data is transferred from batch processing data into streaming data. Speed
of data generation or the processing is similar to “velocity of light” it increases on a continuous
basis. Compared to small data, big data is produced continuously.

Veracity: Veracity is nothing but an uncertainty of data. It refers to data quality and data value.
The best example of veracity is data available on social sites.

18
Value: Generated data from the diverse sources is raw data. Value of the data can be discovered
by analyzing data, recognizing patterns and by predicting behavior.

1.7 Parallel Clustering Algorithm

Most of the optimization algorithms are computationally expensive and involves large number of
verification and iterations. Main aim of parallel clustering algorithms is to complete the task in
minimum time with high relevancy and accuracy. In PCA16 task are carried out in a parallel
manner. Fuzzy C means, K- Means are the famous algorithms used for parallel clustering.

Fuzzy C-means algorithms allow to part of data belong to two or more clusters. This algorithm is
used for pattern recognition and based on minimization of

𝑚
𝐽𝑚 = ∑N 𝐶
i=1 ∑𝑗=1 𝑢𝑖𝑗 ||𝑋𝑖−𝐶𝑗|| 2 ……………..(1.4)
m is greater than 1, uij is a degree of membership for Xi in the cluster J, Cj is a d-dimension center
of the cluster.

Steps involved in Fuzzy c-means algorithm are:

1) Initialize U=[uij] matrix, U(0)

2) At k-step: calculate the centers vectors C(k)=[cj] with U(k)
3) Update U(k) , U(k+1)
4) If || U(k+1) - U(k)||< δ then STOP; otherwise return to step 2.

In first step we initialize membership matrix U, then at K-step will calculate the center vectors C(k)
. In each iteration updations of U(k), U(k+1) takes place. If the difference between U(k), U(k+1) is less
than coefficient of membership function then stop. While implementing all the above techniques
for I-Commerce there are lot of challenges are involved.

1.8 Challenges in I-Commerce

There are enormous challenges are involved depending on the location and diverging from place
to place. Few common challenges are as follows:

The biggest challenge in I-Commerce is security; there is a lack of confidence between consumers
while purchasing products online, consumer need to enter the bank information or personal
information. Where performing this process consumers are having a misconception that while

16
Parallel Clustering Algorithm

19
entering a data, it must be unsecured. Many Misconceptions are there about warrantee and
guarantee of products. It means consumers are always doubtful while purchasing products about
their warranty, whether customer service centers respond to problems or not. Online literacy is
also important while handling products online. Because of computer illiteracy, many peoples are
not able to handle I-Commerce services properly.

Also, there is a myth that online products are not delivered in proper conditions or customers and
if the wrong product is delivered, they will not return the correct product. Consumers are not aware
of the importance of I-commerce. As compared to Physical stores cost is reduced in I-Commerce
as many supply chain phases are removed. Also, consumers are interested to feel the product
experience or personal selection, but that is not possible in I-commerce.

The biggest challenge is in developing countries are not aware of the laws and regulations of I-
Commerce. Because of the unawareness of laws peoples are hesitated to buy the products online.
To solve the above challenges many researchers have done several types of research. Several
approaches have been introduced in I-Commerce to solve the above issues.

1.9 Current Research Trends in I-Commerce

In the generic approach, for product purchasing, consumer needs to visit the store physically and
choose a product with limited varieties. This style of purchasing is a time consuming and wastage
of efforts. Also, user needs to visit the various stores and negotiate the price or compare the prices
for buying products. To solve all these problems, researcher has introduced I-Commerce. Where
buying and selling of the products are available with ease of access with variety of products on a
click for the convenience of customers The delivery of I-commerce abilities directly into the hand
of customers through wireless technology is known as Mobile commerce or M-Commerce. With
the rapid growth of mobile devices, wireless applications and telecommunication technologies
have become increasingly popular. From these technologies, it can be easy to attain the person’s
position through mobile device using GPS service that records the history of user movement. With
a sequence of customers‟ moving logs, it can be easy to know about the moving pathway of mobile
users. Most of the people are using mobile devices in order to purchase items in online through
credit cards. By making the prediction, first, employ the mobile commerce pattern mine algorithm
for effective detection of frequency pattern of mobile users.

20
Due to increase in internet shopping, there is a need for technological innovation, so customer can
select the best product from the best vendor, by using a recommendation system or search engines.
Recommendation techniques are still great challenges in I-Commerce for selection of relevant
products for customer. By applying the different mechanisms in I-commerce, product search
recommendation systems to display the products to the customer by taking the input and traverse
in database. The enormous choices are available for different products with large number of items
with various brands. This creates large database information. It creates a cognitive overload on
customers. I-Commerce space is becoming a very competitive and vendors are finding the various
ways to attract the customer for selling the products online. In order to solve this problem, many
vendors are using several techniques like Amazon uses A9, eBay uses Best match algorithm,
Flipkart using a solar Lucene. However, search engines are not able to solve this problem
completely.

However, preserving privacy has been assumed to be a challenge in I-commerce as the occurrence
of illegal intruders in has subsisted vastly. User information is at a threat of being hacked
effortlessly through the I-commerce service resulting in apprehension and an unsatisfied customer.
Likewise, numerous varieties and ranges of products are available in the I-Commerce applications.
Selecting the right and better product on the basis of the user requirement is perplexing and is a
tiresome task and overcoming these shortfalls necessitates methods to make I-commerce more
user-friendly. Prediction of I-commerce site and products can assist the customers with
opportunities for choosing the providers besides the products. Thus in accordance with these
insights, is aimed at developing a Prediction based Parallel clustering algorithm for recommending
products using I-commerce.

In this chapter, we have emphasized on introduction to I-commerce with different types of I-

Commerce. Also, we have identified different tools of I-Commerce like recommendation systems,
optimization techniques and prediction techniques. I-Commerce dealing with the various
techniques likes big data, parallel clustering algorithms. All these methods alone are not guarantees
that the customer will get the relevant product information in the first search or top of the list. The
required products with different brands are not available with same vendors or different vendors
may not have a similar price. Hence there is a need to develop a transparent model which will
select the best product with relevancy and high accuracy. By considering different feature sets,

21
reviews, ratings, price, priority and extract the products with multiple vendors. In this thesis, we
proposed a model with a primary objective is to propose a prediction based parallel clustering
algorithm using optimization technique for identifying users need and predict the result as per
user’s requirement as well as enhance the relevancy of results.

References and Books

Avasthi S. and Dwivedi A. (2013), “Prediction of Mobile User Behavior Using Clustering”,
International Journal of Scientific and Research Publications, Vol. 3, Issue. 2, pp. 1-5.
Beel, J.; Gipp, B.; Langer, S.; Breitinger, C. (26 July 2015). "Research Paper Recommender
Systems: A Literature Survey". International Journal on Digital Libraries. 17 (4): 305–
338. Dorigo, Gambardella, M, L.M. (1997). "Learning Approach to the Traveling Salesman
Problem". IEEE Transactions on Evolutionary Computation, 1 (1): 214.
Beel, J.; Langer, S.; Genzmehr, M.; Gipp, B.; Breitinger, C. (October 2013). "Research Paper
Recommender System Evaluation: A Quantitative Literature Survey" (PDF). Proceedings of the
Workshop on Reproducibility and Replication in Recommender Systems Evaluation (RepSys) at
the ACM Recommender System Conference (RecSys).
Bratton, Daniel; Kennedy, James (2007). Defining a Standard for Particle Swarm
Optimization (PDF). Proceedings of the 2007 IEEE Swarm Intelligence Symposium (SIS 2007).
pp. 120–127.
Cavazos J., Moss E. B. and Boyle M. (2006), “Hybrid Optimizations: Which Optimization
Algorithm to Use?”, retrieved on 9th March 2017 and from
https://www.eecis.udel.edu/~cavazos/cc-2006.pdf
Chander (2009) “Optimization Techniques” by New Age Science (April 15, 2009)
Evangeline S. J., Subramanian K. M. and Venkatachalam K. (2013), “Survey on Personal Mobile
Commerce Pattern Mining and Prediction”, International Journal of Advanced Research in
Computer Engineering & Technology (IJARCET), Vol. 2, Issue. 12, pp. 3163-3167.
Evangeline S. J., Subramanian K. M. and Venkatachalam K. (2014), “Efficiently Mining the
Frequent Patterns in Mobile Commerce Environment”, Vol. 5, No. 1, pp. 30-39.
Ho, D. T., & Garibaldi, J. M. (2013, July). An improved optimization framework for fuzzy time-
series prediction. In Fuzzy Systems (FUZZ), 2013 IEEE International Conference on (pp. 1-8).
IEEE.
Hong Mee Song et al., 2014; Ali Madadi and Mahmood Mohseni Motlagh, 2014; Pranjali Rathee
et al., 2015; Xianhai Song et al., 2015).
Humphreys et al. (2016), “A Hybrid Optimization Technique Applied to the Intermediate-Target
Optimal Control Problem”, Global Journal of TechnolOptim, an open-access journal, Vol. 7, Issue.
2, pp. 1-8.

22
J. C. Bezdek (1981): "Pattern Recognition with Fuzzy Objective Function Algorithms", Plenum
Press, New York.
J. C. Dunn (1973): "A Fuzzy Relative of the ISODATA Process and Its Use in Detecting Compact
Well-Separated Clusters", Journal of Cybernetics 3: 32-57.
Kandasamy K. and Kumar C. S. (2015), “Modified PSO Based Optimal Time Interval
Identification for Predicting Mobile User Behaviour (MPSO-OTI2-PMB) in Location-Based
Services”, Indian Journal of Science and Technology, Vol. 8 (s7), pp. 185–193.
Kennedy, J.; Eberhart, R. (1995). "Particle Swarm Optimization". Proceedings of IEEE
International Conference on Neural Networks. IV. pp. 1942–1948.
Kim W. (2009), “Parallel Clustering Algorithms: Survey”, retrieved on 9th March 2017 and from
https://s3-us-west-2.amazonaws.com/mlsurveys/46.pdf
Lee S. and Ahn H. (2011), “The hybrid model of neural networks and genetic algorithms for the
design of controls for internet-based systems for business-to-consumer electronic commerce”,
Expert Systems with Applications, Vol. 38, pp. 4326–4338.
Maalini D. and Subashini S. (2014), “An Enhanced Prediction of Subsequent Mobile User
Behavior in Location Based Service Environments”, International Journal of Emerging Research
in Management &Technology, Vol. 3, Issue. 2, pp. 55-60.
Nayak M. and Prakash K. B. (2013), “A Framework for Personal Mobile commerce Pattern
Mining and Prediction”, International Journal of Engineering Trends and Technology (IJETT),
Vol. 4, Issue. 9, pp. 3921-3925.
Niknam T., Nayeripour M. and Firouzi B. B. (2009), “Application of a New Hybrid optimization
Algorithm on Cluster Analysis”, International Journal of Electrical and Computer Engineering,
Vol. 4, No. 4, pp. 238-243. Argade D. and Chavan H. (2015), “Improve Accuracy of Prediction of
User‟s Future MCommerce Behaviour”, Procedia Computer Science, Vol. 49, pp. 111 – 117.
Retail e-commerce sales CAGR forecast in selected countries from 2016 to 2021". Statista.
October 2016. Retrieved 1 January 2018.
Taherkhani, M., Safabakhsh, R. (2016). "A novel stability-based adaptive inertia weight for
particle swarm optimization". Applied Soft Computing. 38: 281–295.
Waldner, Jean-Baptiste (2008). Nanocomputers and Swarm Intelligence. London: ISTEJohn
Wiley & Sons. p. 225. ISBN 978-1-84704-002-2.

23
WEBSITES
[1] https://www.researchgate.net/publication/2955372_The_emergence_of_Mcommerce ,
accessed on 18th September 2016.
[2] http://www.mobify.com/wpcontent/uploads/eMarketer_Optimizing_for_Mobile_
Commerce_05022016.pdf accessed on 26th October 2016
[3] www.sciencedirect.com/science/article/pii/0167819189900367 accessed on 29th
December 2016
[4] www.icommerceteam.com accessed on 26th Jan 2017
[5] http://grigory.us/blog/mapreduce-clustering/ accessed on 2nd Feb 2017
[6] http://www.investopedia.com/terms/e/ecommerce.asp accessed on 6th Feb 2017
[7] Mcom.cs.cmu.edu accessed on 10th Feb 2017
[8] https://www.shopify.com/enterprise/global-ecommerce-statistics accessed on 9th Feb 2019
[9] https://ourworldindata.org/internet#growth-of-the-internet accessed on 9th April 2019
[10] https://www.internetworldstats.com/stats.htm accessed on 9th April 2019
[11] https://www.statista.com/statistics/273018/ number- of- internet- users- worldwide/
accessed on 8th May 2019
[12] www.emarketer.com accessed on 9th May 2019

24