Users’ Behaviour Analysis in Structured E-Commerce Website

This project analyzes customer behavior towards e-commerce websites using data mining techniques. Linear temporal logic model, computational temporal logical model, and probabilistic model checking will be used for analyzing the structure of e-commerce website with the help of web logs. The project aims to understand the fundamental of e-commerce website, how it emerged, why people prefer online shopping over offline shopping, and how it works in the existing system. The project also includes the development of an e-commerce website to analyze customer behavior based on various parameters. The obtained results are discussed in detail. The project uses Java for creating the web pages and Weka tool is used for data preprocessing.

Users’ Behaviour Analysis in Structured
E- Commerce Website
Student Name:
Register Number:
Supervisor:
1

Abstract
The project mainly focused on the analysis of customer behavior towards e-commerce website.
At present, the people around the globe are addicted to online shopping due to its immense
comfort for the customers. E-Commerce has become so common and popular because, it brings
ease into people’s lives. Today’s E-commerce websites are highly effective and powerful to
benefit the customers. The E-Commerce websites provide n number of varieties at one place. It
is understood that e-commerce mainly focuses on customer satisfaction. Keeping all the benefits
of E-Commerce in mind, it is important to know that, it is complicated to understand the
behaviour of each customer, which helps in satisfying the customers. A session of the user on
different websites is stored, which helps in identifying the behavioural patterns of the users.
However, it is possible to capture the information like, what the user/ customer would like to
buy, based on mostly viewed products. The web server stores this type of information in the form
of web logs. The logs helps to analyze the information to know the behaviour of the users.
Therefore, this project works on understanding the customers and their interested products, to
satisfy them. To understand the fundamental of e-commerce website, a deep study has been
carried out. In this part consists of how e-commerce emerged, why people prefer online shopping
over offline shopping, and how it works in existing system. To analyze the customer behavior
towards e-commerce website, three model checking is done. The models are Linear Temporal
Logic model, Computational temporal logical model and Probabilistic model. The stability of
each node’s state is evaluated. An e-commerce website is developed to analyze the customer
behavior based on the parameters Evaluation of products and suppliers, Consumer Buying
Behavior, the customer's need, checking out alternative products and suppliers and purchase
decision. The obtained results are discussed in detail. To develop e-commerce website, Java is
used. To make the website for user friendly, the Bootstrap is applied. The sample Behavioral
patterns are taken and these are applied on Weka tool for data pre-processing. The analysis part
is carried out with comparing three top most e-commerce website and how the people are look
and feel while online shopping. Charts and tables are used in this analysis part for easy
understanding of data.Thus, effective comparison is the key to complete the analysis of this
project. For accomplishing the comparison the help of Web analysts is taken to concentrate on
applying data mining techniques such as k means nearest neighbor (KNN) algorithm for
analyzing user’s behavior. Sequence of action is performed. Also, the behavioural patterns help
to understand the interests of users, to be applied on E-Commerce websites. The linear-temporal
logic model is proposed in this project to help the analysis of the structured e-commerce web
logs. Based on the E-commerce Structure, a common method for mapping the log records can
help in converting them to event logs, which captures the users’ behaviours or actions. Later,
various predefined queries are utilized for identifying the behavioural patterns which considers
various actions performed by the user, during a session. As a final point, the proposed model is
implemented on real case studies of different E-Commerce websites. On the other hand, in the
software development life cycle, the waterfall model is used, which takes care of requirement
gathering, planning, designing, coding and testing. The results of the analysis are provided.
Keywords: E-Commerce; model; Web logs analysis; behavioural patterns; Data mining
2

Table of Contents
Table of Contents.......................................................................................................................................3
1. Introduction-......................................................................................................................................1
1.1 Research Questions....................................................................................................................2
1.2 Aim of the Project......................................................................................................................2
1.3 Objectives...................................................................................................................................3
1.4 Literature Review......................................................................................................................3
1.5 Research Content.......................................................................................................................4
1.6 Dissertation Overview...............................................................................................................5
2. Background........................................................................................................................................6
2.1 Evolution of E-Commerce.........................................................................................................6
2.2 Existing System..........................................................................................................................7
2.2.1 Disadvantages of Existing System.....................................................................................9
2.3 Hypothesis..................................................................................................................................9
2.4 External Interface Requirements.............................................................................................9
2.5 Summary....................................................................................................................................9
3. Related work....................................................................................................................................10
3.1 Summary..................................................................................................................................13
4. Model Checking to Analyse Event Logs.........................................................................................14
4.1 Summary..................................................................................................................................19
5. Software Development Models.......................................................................................................21
5.1 Waterfall Model.......................................................................................................................21
6. Up and Scrap...................................................................................................................................22
6.1 Summary..................................................................................................................................24
7. Implementation................................................................................................................................25
8. Evaluation and Results....................................................................................................................32
8.1 Data Pre-Processing.................................................................................................................32
8.2 Results.......................................................................................................................................42
8.3 Summary..................................................................................................................................47
9. Identifying Users’ Behavioural Patterns........................................................................................47
9.1 Summary..................................................................................................................................48
10. Conclusions and Future Work....................................................................................................48
3

References................................................................................................................................................50
Appendix..................................................................................................................................................51
List of Figures
List of Tables
List of Abbreviation
4

1. Introduction-
The use of E-Commerce websites for shopping ‘n’ number of products has increasing
demand. Along with threats, online shopping has immense benefits that overpowers the
customers. For instance, the primary advantages that overpowers the users include, shopping
from their comfort zone and saving a lot of time. E-Commerce websites has endless varieties and
trending products. E-Commerce websites ensures to satisfy their customers. However, it is a
known fact that, it is complicated to understand the behaviour of each customer, and it varies
from one to another. The sessions of users are stored for recognizing their behavioural patterns,
which helps to capture the answer for the following question-What the user or customer might
wish to purchase? This is made possible only depending on the mostly viewed products. The web
logs are stored by the web server, for analyzing the information to know the behaviour of the
users. Hence, to complete this project, understanding the customers’ behaviour is important.
This project uses Java for creating the Web pages and Weka tool is used for data pre-
processing. More than one model will be applied on various websites to prove the analysis,
which will result into comparison. Thus, effective comparison is the key to complete the analysis
of this project. For accomplishing the comparison, the help of Web analysts is taken, which
ensures to concentrate on applying data mining techniques for modelling users’ behaviour.
Sequence of action is performed. Hence, more complex behavioural patterns are identified by
analyzing the user sessions. Also, the behavioural patterns help to understand the interests of
users, to be applied on E-Commerce websites. The linear-temporal logic model is proposed in
this project to help the analysis of the structured e-commerce web logs. The proposed model is
implemented on real case studies of different E-Commerce websites. In software development
life cycle, the waterfall model is used, which takes care of requirement gathering, planning,
designing, coding and testing.
The importance of response time while searching the products in the websites require in-
depth understanding, to increase the efficiency of E-Commerce services provided by the
company websites. However, comparing our website’s efficiency with other company websites
has a significant importance, as it leads to improvisation of the web services. The recent
advancements in the programming languages and internet, has increased the number of E-
Commerce users. Thus, more reliable, fast and efficient web service is in high demand. In fact,
this helps to attract more number of customers. This investigation contributes to find an effective
1

model that increases the efficiency of searching the products, for the customers/ users. The
response time is used for comparing the efficiency of the websites, in searching the products.
The following are the factors that are considered in the utilized programming languages,
of this research:
1. Portability
2. Versatility
3. Efficiency
4. Java is designed especially for distributed computing with internetworking
capability that is inherently integrated into it. Writing internetwork programs in
Java is used for data transfer in a file. Java has security design as a main part.
Based on Security Java language, compiler, interpreter, and runtime environment
were all developed. Robust means reliability. Java lays a lot of emphasis on early
checking for possible errors, in other languages java compilers is able to detect
Many Problems during execution. Simultaneous Program Execution is known as
Multithreading. In Java is adapted with multithreaded programming, while in
other languages, operating system-specific procedures have to be called in order
to enable multithreading.
5. Java has become a language of choice for providing worldwide internet
solutions ,Why Because its having more properties like robustness, ease of use,
cross-Platform Capabilities and Security.
1.1Research Questions
The following are the research questions of this dissertation:
1) How the customers look and feel about the e-commerce websites?
2) Does the customers prefer online shopping more than offline shopping?
1.2Aim of the Project
The aim of this project is to analyze the customer behavior towards E-commerce website.
The linear temporal logic model, computational temporal logic model and Probabilistic model
checking will be used for analyzing the structure of e-commerce website with the help of web
2

logs. This helps in analyzing the behaviour of the website users. More than three case studies of
E-Commerce sites are considered in this project, to provide clear analysis results. The user’s
behaviours are captured with the stored web logs, which are later converted as event logs, using a
common method for mapping the log records. Then, identification of various behavioural
patterns takes place, using pre-defined queries which the user can likely perform. The researcher
of this project will define the pre-defined queries, to complete the project. Three websites will be
selected to perform real case study.
1.3Objectives
The objective is to analyze the customers’ behavior while shopping through online. This
report will provide an analysis of the three case studies of E-Commerce websites, where
customers’ behaviours will be analyzed. Online shopping website will be developed, where the
user can view the list of products and add them to their carts. To achieve the objective, data
mining will be applied. This will be compared by applying various models such as temporal
logical model, probabilistic decision making model and CLTL model. Throughout this model
checking, the efficiency of the website will be analyzed. Later, data pre-processing will be
implemented using WEKA tool.
1.4Literature Review
According to this paper (Bejju, 2018), In economic development, E-commerce is the
most important catalyst. Very fast growth in usage of web based application and internet
application is decreasing the operation costs of large enterprises. Many organizations have
decided to reconstruct their business strategies, to attain the maximum value and customer’s
satisfaction. E-commerce is not just trading the products; it provides the opportunity to complete
with other gains in the market such as delivering the needs of people within a short time. The
organizations attain business knowledge by using data mining, as it is used to attain the
knowledge from available information in order to help companies to make correct and weighted
decisions. E-commerce ensures fundamental changes to their customers. E-commerce marking
requires e-commerce websites to get a deeper understanding of e-commerce internetworks like
Flipkart, Myntra and Amazon market space. It is used to create interaction, reputation and trust
3

value in developing the business strategies, which supports to build an effective user application.
Also, E-commerce websites are used for company's image establishment, to sell and promote
goods. Moreover, it supports to give full customer support. Basically in an electronic market,
successful e-commerce websites are used to provide success for the company. E-commerce
website’s success is used to reduce the user perceived latency and quality. To reduce the latency,
the extraction of path traversal patterns approach is used in data mining. This approach is used to
predict the user behaviour in e-commerce sites. It also uses the decision tree approach to predict
the online rating of a product. The decision tree approach is used in visualization of probabilistic
business models such as product name, online rating, quantity and product price. In e-commerce
website, to retrieve the website data, extract classification rules are used. The data classification
is used to identify the common characteristics in a website data for categorising them in different
groups.
This paper describes (Sergio et al., 2018) that, In our daily lives, the online shopping is
more and more effective and common. The E-commerce website is very useful for e-commerce
market. To understand the user interest and behaviours is essential for e-commerce, so it leads to
adapt the e-commerce website. In e-commerce website, web server logs are used to store the user
information. To analysis the web server information data mining techniques are applied. The
user behaviours and numbers of the actions are performed by static characterization. This
technique is used to find out more complex behavioural patterns. This paper addresses the
analysis of structured e-commerce web logs by using linear temporal logic model checking
approach. The data mining techniques are used to extract a process model which will usually
provide with either over fitting spaghetti models. This approach is used to represent the event
attributes and types considering the e-commerce web structure and product categorization. It also
navigates the website according to the organization. This paper presents the application of the
approach to the up and scrap e-commerce website. It is used to identify various issues and
improves the organization’s product value.
1.5Research Content
The following are the research areas that will be required for this dissertation- For
Developing website, for implementation of Weka tool and for analysis of case studies on E-
4

Commerce websites, through comparing three other e-commerce website to determine the
customer behaviour. To help developing the website, Java servlet is required for this research.
Then, one must know to work on Weka tool. In addition, waterfall model is considered in
software development and linear temporal logic model, computational logic model, probabilistic
model checking are also the areas to be researched before proceeding with the development of
the website.
1.6Dissertation Overview
Chapter 1 is the introduction part for the whole project. It signifies the problem and aim
of the project. This chapter also contains Literature Review. This section contains reviews of
various researchers that enlighten with additional information about the topics to be discussed in
this dissertation.
Chapter 2 briefly sketches the background for the literature Review presented in the
dissertation. This chapter acts as a base for this project, to understand the concepts involved in
the project. The hypothesis of the project are presented in this chapter.
Chapter 3 furnishes details about the related information for this project. The other areas
covered in this chapter includes- Data mining techniques, web usage mining , web transaction
and web page clustering technique and explanation of Markov model.
Chapter 4 provides in depth information on model checking for analyzing the event logs.
Chapter 5 comprises of solution and validation part. This chapter brings the effective
solution and convincing validations for the project. It is provided with a test plan to determine
the solution.
Chapter 6 highlights the strategies of up and scrap, which are listed and discussed. Then,
the implementation of statistics mining strategies on access logs is unveiled. As a whole, this
chapter discusses about the web statistics mining.
Chapter 7 is assigned for evaluation and results, where the output is analyzed. Further,
this chapter concludes with the achieved results.
In Chapter 8, identifying all the behavioural patterns of the users are studied.
The last part of this dissertation is entitled as Chapter 9, which includes the conclusion
for this project. All the areas covered are discussed in this chapter to express the purpose of this
project.
5

2. Background
2.1Evolution of E-Commerce
The retail business throughout the world has been revolutionized in the last two decades.
It has evolved from scratch and has passed through various significant milestones, which brings
us to today's E-Commerce. However, E-Commerce during its initial days, was least popular and
least used. People were scared to opt it and the retailers were scared to venture it, due to n
number of threats. The factors that affected the E-Commerce business are- The customers were
not able to touch and feel the product that they wish to purchase. This particular factor had a
greater negative impact on the E-commerce business. The other major concern was secured
payment mode, for their purchase. However, the customers also doubted the trustworthiness of
online shopping.
The development of internet and technology, the website services are enhanced. The
reason for the growth of E-Commerce is regarded as connectivity, through internet between the
customers and businesses that websites. The customers have easy access to their offered products
where they can see, compare the products and buy them with just some clicks. For increasing the
number of customers, the online retailers are in quest to improvise their e-commerce potentials
that maximizes the website’s responsiveness. The customers of this generation are quite savvy,
because they have complete details about their interested product, by means of reviews and
product details furnished in the websites.
Thus, purchasing from a particular website completely depends on the customer.
However, the online shopping behaviour of the customers can be studied, to help the customers
find their interested product. Generally, in this internet oriented world the customers tend to surf
internet before purchasing any product, where the mode of purchase could either be through
online shopping or by manual shopping. In terms of customer behaviour, the online shopping is
quite similar to manual shopping.
6

Figure: Expected E-commerce Sales worldwide till 2021 (Orendorff, 2017)
2.2Existing System
The current E-Commerce website mostly use an under fitting flower model or process
oriented model or the spaghetti model. Today’s data mining tools and techniques have same
constraints based on analysis survey on the other hand; they don’t work in a direct way for the
sequence of events (click stream, data associated to each click).
Assessment of merchandise and providers: Now the time to offer the capability purchaser
with demonstration products, visits to present customers and plant visits or 0.33 party
testimonials. And also special pricing programs or stocking incentives can be taken.
Ready to area an order:
That is the time for private touch. It helps to reach customers easily.
Evaluation of product and supplier overall performance:
For selection of shopping, extra reassurance is required by the consumers. Evaluate
conferences
7

and helpline guide to provide reassurance properly after income help and continue exposure to
marketing and press coverage - justifying the acquisition decision.
Observe on purchase:
The primary purchase should no longer be visible as the cease of the process, however as the
starting of a protracted-term enterprise relationship.
Consumer shopping for conduct:
There are numerous fashions of patron shopping for behavior, however the steps below
are
pretty commonplace to maximum them.
The purchaser identifies a want:
This is often initiated by means of PR coverage, which includes phrase of mouth. The
Consumer might also have seen a pal or movie star using a service or product, or cognizance
may also had been sparked off by advertising and marketing.
Searching out facts:
At this degree, the customer needs to know more and looking for data. Advertising
and PR are nevertheless essential, however product demonstrations, packaging and product
shows play a significant role. Now the time to set up your sales employees and the customers
find motion pictures and brochures which are useful.
Finding out opportunity products and suppliers:
The consumer is now looking to select between products, or company up on the purchase
decision. This Platform is used for promoting product guarantees and warranties, and
maximizing packaging and product displays. Sales personnel greatly have an effect on the patron
at this level and income promotion offers emerge as of interest. Unbiased sources of facts are
nevertheless of hobby, which include product check evaluations.
Buy choice:
That is the time to 'tip the stability'. Income promotion come by their own, and if
appropriate,
income force incentives wants to make sure that your sales employees are incentives to shut the
deal.
Using the product:
8

The Cause of Expensive purchases is called cognitive dissonance - a fear that the
consumer
has no longer made the right selection. Main process is to reassure the customer with the aid of
offering precise patron care, simple guidance manuals and loyalty schemes. Still it is to be
exposed to testimonial advertising and marketing to reassure that they did the proper decision.
2.2.1 Disadvantages of Existing System
The disadvantages of the existing system includes:
1) It costs more for competing the prediction of the users’ behaviour methods and to
implement data mining techniques.
2) Log based analysis is not much accurate and tag based implementation may result
in high download time of the website, privacy concerns and so on.
2.3Hypothesis
The hypothesis are listed below:
H1: How the customer behavior analysis can be done through model checking?
H2: How the people look and feel about e-commerce website?
H3: Is the datamining technique really helps to determine the behavioral patterns?
2.4External Interface Requirements
The external interface requirements are listed below:
1) User Interface: It provides a user-friendly Java Graphical User Interface. Java
swing is also used. Bootstrap, HTML5 and CSS.
2) Hardware Interfaces: The Java capabilities are made by giving the interaction
between the user and the console.
3) Software Interfaces: JAVA1.6. Netbeans IDE are required.
2.5Summary
This chapter emphasizes on the background of the project. The literature review in the
previous chapter benefits by the researched background in this chapter. The evolution of E-
9

Commerce is studied and then the existing system with its disadvantages are presented to have a
broader picture of why this research is carried out. At last, the external interface requirements are
represented. It is observed that, from the background effective hypothesis questions are identified
to be worked on during the overall research of the dissertation.
3. Related work
The ability to discover the wealthy customers of their shopping for decision making
procedure is heavily motivated by way of the insights won in the second before the buying
choice is made. Consequently to build a predictive choice making solution in the direction of a
particular product or services might be very powerful to boom ROI, conversion and client
satisfaction.
The method of click stream records is distinct than regular CRM records processing to
construct the predictive version. The information is typically semi-based and accumulated from
respective internet server. The under proven discern depicts the special grade by grade system to
carry out the web consumers’ conduct evaluation.
Lot of customers view e-commerce sites, but most effective few lots site visitors may
additionally purchase the products. The E-retailer wants to enhance the client revel in and would
love to enhance the conversion rate. The Task objective is to get the shoppers ability primarily
based on their demographics, historical transaction pattern, clicks pattern, browsing pattern in
exceptional pages and so on. Records evaluation revealed the buyers’ purchasing behaviours
which are quite related on their range of clicks, duration of period, previous session, purchase
session, clicks price consistent with consultation and so forth. The use of the answer framework
(discern 1) and making use of facts mining and predictive analytics techniques, the propensity to
conversion rankings of each traveller has been derived. This results in more than one blessings of
the e-store to offer proper & centred product for the customers at right time, boom conversion
charge and enhance customer pride. Based totally on this evaluation store can optimize the
advertising method primarily based on diagnosed hidden elements of conversion and understand
the acquisition funnel of clients.
10

In E-commerce, process the server logs to extract the sequence of user navigation events
are done by data mining techniques. The navigation events are not directly mined. High level set
of data summarizing is one of the characterization of data mining techniques, which happens
during the user navigation.
The data mining techniques are used for the web usage mining. The web usage mining
has two types of clustering methods which are web transaction clustering and web page
clustering. The web page application is successful of the adaptive website. The web page
algorithm is known as pageGathe. The Proposed algorithm is based on the partitioning group of
the Web pages to synthesize index pages is do not exit initially. According to the navigational
histories ,the user access interest is represented by index pages. The example of result is
clustering user rating which successfully uses data preparing step and improves the
recommendation of the scalability using k-Nearest –Neighbor(kNN) algorithm.
11

The web transaction and web page clustering technique represents employing, using the
traditional k-means characterize user access the clustering techniques, based on the data mining
web usage data of web personalization (Mikut and Reischl, 2011).
12

These techniques have been proven to have efficient form of experiment result. The usage
pattern is explicitly captured at the user session of the level. The web page segments have
various types of the web data. The web data is utilizing called probabilistic semantic latent
analysis (PLSA) model. This model is based on the principle of statistics. The result of web
proposed LSI based LUI generates user profiles for web transactions. The relationship of the co-
occurrence of user sessions and usage data model has session page matrix. The employ is based
on the SVD algorithm. This algorithm captures the information form by partitioning user
sessions. The k-means cluster algorithm is generated for the user session cluster. The user access
to the web page has a pattern. various tasks of behavior oriented set with page-weight pairs are
created by user profile. Each weight pairs reflects the significance on the web page. The
experiment’s result has two real world datasets for validating the effectiveness and efficiency of
the LUI algorithm. The result provide user access pattern and reveals the underlying relationship
among the user visiting records. The researchers apply mining technique, which predicts the
behavior of the user. The navigational user sequences create statistical and probabilistic model to
predict the user’s next click. This model is called as Markov chains. The process of creating the
Markov model is very expensive. This type of model information is used for very short term but,
the current navigation state can reach long term goals. The algorithm of Markov chains improves
the predications of these statistical models. The group of user sessions applies some clustering
algorithm and later generates a specific Markov chain for each of the obtained cluster. In
comparison, the clustering approach is a commercial tool and has advantage of mining technique
which provides causal relations with users and traces global view for the whole session (Kudyba,
2014).
This framework can be customized with the unique requirement, for every client. Without
editing the present procedure and infrastructure the era agnostic nature included any current
database platform. In addition, the data can be sliced and diced, the use of our proprietary facts
mining algorithms and strategies can lead to the clients in the direction of optimizing their
method. However, it can also extensively improve the ROI and growth of consumer pleasure.
13

3.1Summary
The chapter enlightens with the importance of data mining techniques in websites. It is
scrutinized that the web usage mining performs two types of clustering methods such as web
transaction clustering and web page clustering. It is determined that the web page application
called as pageGathe is successful for the adaptive websites. This chapter discusses about data
mining techniques, web usage mining, web transaction, and web page clustering technique and
Markov model.
4. Model Checking to Analyse Event Logs
Quantitative verification system model with stochastic behaviours by using Stochastic
model checking. At present, stochastic version checking makes an incredible development. In
this venture, the high degree machine model is used, because the formal version for machine
with stochastic behaviours; Based on probability bounds PCTL* is selected as the assets
specification compare than PCTL and LTL. Then, the PCTL* stochastic version checking
algorithm for LPPN (label-prolonged probabilistic Petri internet) is applied.
Algorithm Process of PCTL model
14

The probabilistic accommodation authoritative adaptation is acquired for bearing the
heuristic selections at every akin of the agreement arrangement via the acceptance of the
academic accommodation timberline situation. Probabilistic best authoritative action is primarily
based at the accomplished agreement accompaniment facts that formulates the multistage
Markov alternative botheration on the achievability amplitude ⟨𝑆, 𝐷, 𝜋, 𝑝, 𝑅, 𝐻⟩ as
accidental parameters.
Acquiesce𝑆∈𝑠1, 𝑠2,...,𝑠𝑛 be the bound set of agreement country spaces acquired
throughout the bound agreement method, 𝐷∈𝑑1, 𝑑2,...,𝑑𝑛 the bound set of choices are
accessible at anniversary nation space, and 𝜋∈𝜋1, 𝜋2,...,𝜋𝑛 the bound set of negotiation
guidelines (patterns) empiric via the contributors agreement method. Any agreement advantage
𝜋𝑡: 𝑠𝑡 → 𝑑𝑡 is the mapping that selects an adapted best 𝑑𝑡 in the advance of the agreement
nation 𝑠𝑡. Then, 𝑝 is the accompaniment alteration appropriate that defines the achievability
administration as action denoted as 𝑅: 𝑆×𝐷 → 𝑟. The strategic decision making diagram is
shown below. It represents the relationship among Action, Tactics and strategy.
15

Figure 1 Decision making strategy
Then, the compliments (𝑠𝑡, 𝑑𝑡) provides the accolade amount acquired by agency of the
abettor agent, afterwards acknowledging the alternative 𝑑𝑡 fabricated in the agreement nation
𝑠𝑡.Finally, 𝐻∈ ℎ𝑡10 , ℎ𝑡21 ,...,ℎ𝑡𝑛𝑛 represents the set of all country alteration annual within all
through the past negotiation states transitions at time on the spot𝑡∈𝑡1, 𝑡2,...,𝑡𝑛.
An agreement action accommodation at anniversary commonwealth is probabilistically
modeled in phrases of codicillary adventitious administration for accepting boundless
anticipation of agreement aftereffect with none adverse result. At any time, actual 𝑡∈𝑇, the
agreement adjustment begins in any one of the negotiation states 𝑆∈𝑠1, 𝑠2,..., makes absolute
best central academic best timberline book alive for the next state. Appropriate here, the
accommodation time aeon 𝑑1, 𝑑2,..., represents the multinomial best variables allowance in the
nonterminal. The actualization of probabilistic decisions at the nonterminal is authentic as
follows:
p ( dk
1
sx
, sx ) ,
p ( dk
2
sx
,dk
1 , sk
1
),
16

p ( dk
3
sx
,dk
2 , sk
2
) ,
.
.
.
.
p ( dk
T
sx
, dk
T−1 , sk
T −1
)
𝑘denotes the able best 𝑑𝑥∈𝐷 taken over the corresponding accommodation timberline
apropos the negotiation date 1, 2. . . 𝑇. Then,0𝑥, 𝑠1𝑥,...,𝑠𝑇−1𝑥 denotes the apparent agreement
nation of the corresponding agreement akin 0, 1, 2. . . 𝑇 − 1. In addition, the probabilistic
choices fabricated aural, where the terminals can be represented as apparent in the afterward
equation:
p ( s y
sx
, dk
1 , dk
2 , dk
3 , … .. , sx
T−1
)
The broking assistant (player) agreement action makes the alternative of accommodation
𝑑𝑘∈𝐷 that, actually relies aloft on the abreast agreement nation 𝑠𝑥. An alteration achievability
from the agreement states 𝑠𝑥 to 𝑠𝑦 at any time 𝑡∈𝑇 describes the set of commonwealth
alteration affairs as declared central of the afterward equation:
psx → sy
d k
= ∑
y=x+1
n−1
psx → sy
d k
, ∀ sx ∈ S , ∀ dk ∈ D
𝐷× 𝑆 → 𝜋(𝑆). The accompaniment alteration befalling (𝑑𝑡, 𝑠𝑡, 𝑠𝑡+1) denotes the
agreement accompaniment alteration from the modern-day commonwealth 𝑠𝑡 to the new
country 𝑠𝑡+1 based at the adjudicator alternative 𝑑𝑡. Action denoted as 𝑅: 𝑆×𝐷 → 𝑟. Then,
the praise characteristic (𝑠𝑡, 𝑑𝑡) presents the acclaim amount acquired with the aid of the
broking agent, afterwards acknowledging the best 𝑑𝑡 fabricated aural the negotiation
commonwealth 𝑠𝑡. Eventually or later, 𝐻∈ ℎ𝑡10 , ℎ𝑡21 ,...,ℎ𝑡𝑛𝑛 represents the set of all
17

accompaniment alteration history determined at some date in the past negotiation states
transitions at time on the spot𝑡∈𝑡1, 𝑡2,...,𝑡𝑛.
Count on that, the summation of all nation transition chance is same as 1; and∀𝑠𝑥,
∀𝑠𝑦. The transition probability is 0≤𝑝𝑠𝑥 → 𝑠𝑦 ≤ 1. The choice characteristic 𝑑𝑘 is associated
with the country transition from 𝑠𝑥 to 𝑠𝑦 gets the following rewards as 𝑟𝑑𝑘𝑠𝑥 → 𝑠y. The goal of
this probabilistic decision making hassle is to discover the sequence of suitable choices that
maximizes the entire anticipated reward 𝑅. It’s far depicted within the following equation:
The alternative maker’s (broker agent) accolade acquired at some date in the alteration
from nation 𝑠𝑥 to nation 𝑠𝑦 beneath the alternative 𝑑𝑘. The accepted atom accolade
appropriate of this alteration is declared as follows:
Let the braking’s agent be the state transition opportunity price obtained for the selection
𝑑𝑘. The transition reward is formulated for computing the reward function in terms of
computational values as proven within the following equation:
r sx→ sy
d k
Assume that the abettor agent agreement action receives the accolade ethics 1, 0.5, and 0
from the adversary denoting the acceptance, adverse proposal, and rejection, respectively.
Explanation of Model Checking analysis done on UPPAAL tool
To do model checking analysis for e-commerce website, UPPAAL tool is used. It is an
executable jar file. It should be installed in our system. The implementation of model checking is
explained in step by step. The main objective of model checking of e-commerce is to identify the
reachability of the designed system. Reachability of the system can be defined as the any state of
the system should not allow deadlock when the process execution.
To prepare the event logs, of the website, the model checking is needed for the
application. In this model checking process, we should analyze the website systems two major
18

components that are provided as the web services with REST protocol. Firstly, the UPPAAL tool
generator allows us to upload csv file and it transforms as the input log file. IN second step, the
model checker investigates the previously loaded file. The SPOT library files are used in the
model checker to analyze the LTL model checking process.
The above given figure shows that the starting window of the UPPAAL tool.
Moreover, the model checking tool has more advantages rather than other temporal logic
formulas. Because, in this tool, we can define and declare set of variables and macros in an
easiest as it provides the possibility of writing LTL formulas quickly. In a subset, there are
multiple ways to define the variable. One of them is ‘Enumeration’. In this way, the variables are
defined as a range of identifiers or regular expressions. If a set named VAR is defined once, the
LTL formula will be applied on each variable belongs to the set VAR. This tool automatically
checks the set VAR. Another way to define variable is ‘Macros’. Generally macros can be
defined with the elements of logical OR or AND operation and performed at all elements.
Another way is to define along with the declared name and to avoid of writing same formula
more than a time.
19

The most of e-commerce websites and web pages are developed with more interactive
manner and it helps their customers to navigate into web page interactions. In this approach,
GET operation is performed to retrieve required data and the POST operation is performed for
getting information about the product, buying the product, sign up, sign in and so on. These
operations are the frequent interactions on these websites.
On e-commerce websites maintains web logs. In these web logs related information
regarding IP accessed by the computer, location of the user, time and date of access. How many
times the user logging in to the website and how many times they have purchased products rather
than other customers. These information are required to analyze to achieve the objective that is
analyzing the behavior of the customers towards e-commerce websites.
So that the suitable model checking method should be chosen to analyse theevent logs of
the website. In this analysis the product categorization can be classified as per the event logs.
The website structure of the developed e-commerce website is shown below. It is designed as the
tree structured one. The movements of web pages at each level such as Level 0, Level…Level n-
1. These meaningful set of event logs is used to find the behavioral patterns with the help of
model checking process.
20

The typical structure of the e-commerce website for organizing the selling products is
shown above. It is explained below.
It starts with the Home page (degree zero) and it is the main section of the developed
website pages. In this page special styles are applied.
1) Main sections, the products or categorized and organized in these page. The users
have the accessibility to access this page and shop any product with required
quantities. In most of the e-commerce websites, the unique products are categorized
in an individual shopping section. But it is not compulsory for all websites.
2) Secondary sections: The signing up and signing in pages are considered as secondary
sections of the websites. The main objective of these pages are providing the
accessibility permission to the authorized user to access the products. If the user is not
authorized, they could not access the products and purchase their orders.
Therefore, we should able to differentiate the web pages based on these two sections.
After completing the authorization process, the product can be accessed by the customer
otherwise, they could not make an order and they can view the product list in a temporary
brought to the session. For example, the user can view and add the product into wish list but they
could not purchase them. But while adding them into wish list is also a temporary session. If the
user is authorized then the link will get into temporary session where the pages include product
cart, summary, payment and so on.
21

After stimulating the generated model on UPPAAL tool, the verification of reachability
is done and shown below.
The probabilistic decision making model has been checked and summarized below.
Probability Matrix
22

Verification Process of Model checking
Thus the reachability of the designed model is resulted as ‘Yes’ as given in the above
screenshot.
4.1Summary
The overall idea is to make the complete use of the internet log records and the usage of
statistics mining applications. Data mining is aimed to find out the requirements, shape, content
material, utilization styles and so forth of customers and internet pages.
Figure 2 Model checking analysis and result of web logs
23

Sensible internet caching algorithms are the equipment to predict the web requests. “these
internet-caching algorithms are capable of adapt their behaviour on the premise of the user
access styles, which in step are extracted from the historic access records recorded inside the log
documents with the aid of information mining strategies.”
The aim of these algorithms is to increase the quantity of internet pages which can be
retrieved directly from the cache as a substitute inquiring for them from the server. There are
numerous web caching algorithms, where three of them are listed: common patterns inside the
case of frequent patterns, we extract from the web logs the styles that follow the form.
A→B (if A, then B). If A has been asked then B is in all likelihood to be requested next.
Decision trees inside the case of selection timber, we develop a decision tree in the foundation of
the ancient facts within the internet logs. But in this situation, concentrating on the time is
required till the subsequent request.
Page gather - This set of rules use the information mining clustering to group a set of
web sites which are regularly visited close to each other and distant from other organizations.
Records mining clustering differs from the traditional clustering in a way that it is able to area
one report in a couple of overlapping clusters and questions where these kind of records can be
accumulated.
The whole thing you do even as surfing inside the internet can offer useful statistics, as
an example, for website online designers.
Clickstream information is the path that the user creates whilst steering through the sites
and by following the hyperlinks. It can be used to assess the visitors and reputation of the page
buying chart, then it can provide information in e-business, in which the purchases had been
made and where the purchaser left the order unfinished.
Psychographic statistics would encompass records on user’s attitudes closer to topics,
products and so forth. Buying behaviour and ideals. Get entry to statistics counts the time
between the final and next get entry to the same URL.
Time statistics offers records on quantity of time a consumer spends exploring the web
page, the product or topic he or she is interested in. All that offers us with the beneficial records
about the users but how a long way are we able to pass till it becomes a privacy subject? Is it
appropriate to document all of the user activities with the intention to discover how users
perceive the web site?
24

These valid analysis can be done through the probabilistic model checking. When
comparing to other two models, it shows high probability of steady state at each iteration. Thus
probability model checking produces efficient behavioural model to analyse the customer
behaviour.
5. Software Development Models
5.1Waterfall Model
Business Requirements – on this level, the product control institution comes up with
requirements for the software program. All necessities are grouped into a document that's the
input for subsequent level. Say requirement can be – capability to go into new orders inside the
gadget, while the order is entered, the system must auto-generate the order number.
Functional Design – this would difficult on the necessities. So right here display screen
mock-ups are made, interfaces/integration points with different structures are recognized, open
APIs are recognized, reviews, workflows et al are also etched out. So this successfully turns into
the bible for the developer to begin the design. Additionally in this document are referred to, the
product use cases and useful test instances.
Technical Design – beginning with functional specifications, developer is going on to
design the UI pages, business common sense, API, reports and so forth. And ultimately comes up
with a document. And unit take a look at cases must also be diagnosed and documented right
here.
Code –It’s just the problem of converting them right into a practicable code. So once
more the improvement crew pitches in and cranks out the code.
Unit Test– The developed system has many modules and these are individually tested.
25

System Test – After completing unit test, the small modules are integrated and then the
entire code package is tested.
6. Up and Scrap
The behavior of e-commerce understands the customer’s needs and also company need
understand what motivates them to purchase and services are on the shopping list. The up and
scrap is leader of business in Spain for the product of sale equipment for scrapbooking. It was
established in 2012, in Spain and more than 25,000 clients performed in over 85,000 shipments.
The up and scrap website includes the more than 2,500 products. It can be purchased and
accessed using from many sections. The figure below shows the structure of web home page in
website (Quirchmayr, 2012).It is used to map for business and develop to motivate the company.
The many way of website access this company. The given below categories are using the
developing the company.
Strategies of up and scrap
Marketing mix
It is marketing achieve goals for selling and buying promoting products and services
(Hernandez et al., 2018).
Marketing strategy
Marketing mix with product element defined
Four p’s of marking
 It is product and product perception.
 Customer value or Custer benefits
 Promotion and distribution.
Main section of up and scrap
Paper
The scrapbooking of shopping chart is provide information in e-business from where the
purchase and were made. And also this section is provides the lot of product is how to perform
and information. This includes two subcategories (Jin and Lin, 2012).
Decorate
The decorate is used to provide the information such as wooden and metal ornaments,
buttons, washi tape, ribbons, pearls and thicker etc.
26

Stamp
It is used to require the product of stamp and paint are offered and including the different
inks set of stamps.
Tools
The main tools are scissors, glue, cutter, and tape.
Project life smash
This used to our own albums and photographs and memories (Siau, 2015).
Home decor-diy
It is using to personalize in decorate your home.
Gifts
The gifts provide the starter kits and gift vouchers.
Secondary sections
The secondary sections function is new products and special offers and it is include the
multiple criteria (Oatley and Ewart, 2011).
This categories includes the two main function are given below
 Section with temporary products
 Section with permanent products
Section with temporary products
Offers
This section provides the new offers to client or user. The users are directly to see the
offer in the product.
New products
The new products are added in the list of scrapbooking.
Section with permanent products
This section includes the brand, thematic, collections and designers.
Brands
These categories provide the information of the all product brand. It is using to user
choose the better product. Currently 151 products are available.
Thematic
The thematic is according to the different themes like summer, babies, dog wear etc.
currently 58 different thematic are available.
27

Collection
Function of collection is all groups of products and new products are collected. Currently
five hundred forty-three collections are available.
Designers
This category allows the user to see the product and satisfaction of the favorite designers.
It is help the user can choose the product in our choice.
The website structure designed in all the products are reach to the user in following many
different navigation paths (Effective Dynamic Pricing for E-Commerce Businesses by
Estimating Customers Behavior Using Data-Mining, 2017).
6.1Summary
Web statistics mining is important strategies that correctly manage the responsibilities of
looking the needed facts from the net source, improving the structure of web page to offer higher
internet carrier quality and coming across the informative information from the internet for
superior internet packages. Internet information mining may be labelled into three sorts of
internet content material, web structure and web utilization mining. On this have a look at, we
consciousness on internet utilization mining: that is, coming across consumer access sample
understanding from internet log documents, which comprise the historic touring statistics of
customers at the internet site. Web content material Mining: it is an automatic technique that
goes past keyword extraction. For the reason that content of a textual content file affords no
gadget readable semantic, a few processes have suggested restructuring the record content in an
illustration that could be exploited by using machines. There are two companies of internet
content material mining techniques: those who directly mine the content of documents and
people that improve at the content material search of different tools like search engines. Web
structure Mining: worldwide internet can screen extra statistics than just the facts contained in
documents internet usage Mining: internet servers record and accumulate facts about person
interactions every time requests for resources are acquired. Analysing the internet get entry to
logs of deferent internet websites can assist apprehend the person behaviour and the web
structure, thereby improving the layout of this significant series of resources. There are
predominant inclinations in web usage Mining pushed by using the applications of the
28

discoveries: fashionable get right of entry to sample tracking and customized utilization
monitoring. The overall access pattern tracking analyses the internet logs to recognize access
styles and trends. Those analyses can shed mild on higher shape and grouping of aid vendors.
Many internet evaluation tools existed but they may be limited and normally unsatisfactory.
We’ve designed an internet log facts mining device, blog Miner, and extra strategies for using
statistics mining and online Analytical Processing (OLAP) on dealt with and converted internet
get admission to files. Applying statistics mining strategies on access logs unveils exciting access
styles that may be used to restructure websites in a greener grouping, pinpoint effective
advertising places, and goal unique customers for particular selling ads. Custom designed usage
monitoring analysis individual trends. Its reason is to personalize web sites to customers. The
statistics displayed the intensity of the web page shape and the format of the resources can all be
dynamically custom designed for each person over the years primarily based on their get right of
entry to patterns. even as it is encouraging and exciting to look the diverse potential programs of
web log record analysis, it's far vital to recognise that the fulfilment of such applications depends
on what and how much valid and dependable expertise one can find out from the big raw log
facts. Modern internet servers shop restricted data approximately the accesses. Some scripts
custom tailored for some websites can also store additional information. But, for an effective
internet utilization mining, a crucial cleaning and records transformation step before analysis
may be wanted.
7. Implementation
The project starts with the index.html page as in the below screen shot.
29

Then right click on the index.html. There will be an option Run File. Click that option.
Then the project starts running by showing the message ‘Starting Browser’ on the left
side down.
30

The output screenshots are given below.
Home Page:
The name of the shopping website has been given as smart shop. The Index page is
designed as below figu
To attract customers, the new arrivals, special offers and discounts are highlighted in
Home page. It helps to sale more products by the customers.
31

As the consumers are looking for trendy products, these are shown in Home page itself.
What is the actual price and special discount offers on its prices are displayed to make this
website as efficient to improve the shopping experience of customers.
32

The Menswear products such as clothing, wallets, footwear, watches and so many
product are listed with detailed description.
To make the website convenient to choose products for the customers, it is facilitated
with filters. The consumer can view the products by filtering them by its price range.
33

The new electronic gadget collections are shown in electronics.html page.
The favorite products can be added to their carts and it will be purchased later. After
adding the item into the cart you can see that in cart items.
34

In Check out page, there will be shopping basket box. In that, the total price of the
products that are added to the cart will be shown.
35

8. Evaluation and Results
Data pre-processing is the first step of web mining usage analysis. the raw data have
relatively low business value when they cannot be transformed and processed to produce
actionable knowledge. The goal of this analysis is to improve the Up and Scrap website design.
8.1Data Pre-Processing
Data pre-processing is one of the important process in data mining which is used to
convert the web logic into some complete, accurate and reliable data source based on the data
mining algorithms. In data mining process 60% of the work is based on data mining. Because the
data’s are incomplete, in accurate with noise. Hence data mining is required to improve its
quality. It also saves the time, space and it plays a vital role in decision making process. This
data pre-processing includes the following processes.
1) Data cleaning
2) Data integration
3) Data transformation
4) Data Reduction
Data Cleaning
The users log file usually contains unwanted data and it creates a negative impact in data
analysis process. Hence it can be deleted through the following ways.
Filling the Absent Values
a) Filling the missing values with average sample values.
b) Filling the missing values with average of all sample values.
c) Filling the missing values with global constants like "unknown"
Smooth Noisy Data
Beginning
First we have to smooth the data using equal frequency, then use the bin median and then bin
boundaries.
Clustering
This method is used to cluster the similar values to get some isolated data.
36

Data Integration
Data integration method is used to merge the data from different sources to build a data
with consistent form. But this process has some problems. The first problem is it compares the
two different attributes from different source. Next problem is reduction. If some property is
derived by one or more attributes then the difference between the attributes leads to reduction.
This reduction can be examined by different methods.
E- Commerce websites deals with complex data from the web to e-commerce domain.
Most of the e-commerce organizations insights the web usage mining to improve their processes.
Data pre-processing is the very first step of mining process. To enable the data analysis process,
the raw data with user logs are pre-processed to remove the uninteresting requests and it
identifies the sessions to enable the analysis. These web logs are in Common language and it has
raw information’s like IP address which is used to initialize session and time requests.
This data pre-processing step is carried out in three steps. First two steps are common for
all the mining process and the final step is used to prepare the logs and it checks the model
techniques (Poggi et al., 2013).
Cleaning User Logs
1) In this process the unwanted records which distort the result are removed through the
following steps.
2) It removes the automatic requests which is sent by spiders, crawlers and robots. The IP
address corresponding to those robots are removed by the automatic tools.
3) Requests which has error status codes also deleted by navigating the errors.
4) It deletes the requests which has multimedia data. These requests are automatically sent
by the browsers.
Identification of the User
The main goal of this phase is the same session events are grouped. For that it uses,
heuristic in which the user session is created with the same IP address of those events. This
values are used in web usage process to analyze and observe the behavior of the logs.
Preparing Log
37

In this process the log file is prepared which is used in the model checker. It has two
types of actions. Categorization is the first step in which the records are analyzed to identify the
high level events and it extracts those information. Simplification is the other process in which
the contents of the logs are reduced to increase its effectiveness. By the analysis of CLF logs
different events are identified . It automatically classifies each events and it considers only POST
or GET requests by analyzing the URL.
The GET requests are classified by the correspondence of main section or secondary
section. The remaining requests are identified based on the different types of events with their
specific URL. These events has different actions and they affect the different parts of the
website.
The remaining requests are identified with different events based on the unique URL.
Different types of events affects different sections of the website. It also provides superfluous
information from the analysis process. In the final log file different types of information are
recorded with event name, request IP address, time stamp , requested URL, type of operation
whether POST or GET and , HTTP status code. But we can’t identify the category and
subcategory of the products which is belongs to web server logs. To automatically obtain these
information structure mining technique is used. The logs are analyzed in the categorization
phase and then simplification process is performed. This phase is used to reduce the amount of
information which is added in the log file. When removing the duplicate records, the amount of
log file is reduced by 31.93% from the previous log file.
The customer behaviour data is used to improve the customer retention. This data set is
used to help user to predict the customer behaviour to retain the customers. This data set also
used to analyze the all the relevant customer data and develop focused customer retention
programs. The data set is shown below.
38

The Data set includes the following information,
1) Customers demographic information (Gender, age and if they have dependents
and partners)
2) Customers account information (Contract, paperless billing, total charges,
monthly charges and payment method)
3) Customer’s services have signed up for interinternet, online backup, online
security, tech support, streaming TV and movies, multiple lines, device protection
and phone.
4) Customers who left within the last month.
Data Pre-Processing Analysis
Here, we are using two types of classify algorithm to find the standard deviation, mean
and number of error. The algorithm is listed below.
1) Naïve Bayes
2) Decision table
The naïve Bayes algorithm is used to find the standard deviation, mean and so on by
using cross validation method. The decision table algorithm is used to determine the number of
error in the model. It use the percentage split method. These two algorithms are discussed in
below.
39

Naïve Bayes Algorithm
40

gender
Female 2550.0 940.0
Male 2626.0 931.0
[total] 5176.0 1871.0
SeniorCitizen
mean 0.1287 0.2547
std. dev. 0.3349 0.4357
weight sum 5174 1869
41

precision 1 1
Partner
Yes 2734.0 670.0
No 2442.0 1201.0
[total] 5176.0 1871.0
Dependents
No 3391.0 1544.0
Yes 1785.0 327.0
[total] 5176.0 1871.0
tenure
mean 37.57 17.9791
std. dev. 24.1114 19.5259
weight sum 5174 1869
precision 1 1
PhoneService
No 513.0 171.0
Yes 4663.0 1700.0
[total] 5176.0 1871.0
MultipleLines
No phone service 513.0 171.0
No 2542.0 850.0
Yes 2122.0 851.0
[total] 5177.0 1872.0
InterinternetService
DSL 1963.0 460.0
42

Fiber optic 1800.0 1298.0
No 1414.0 114.0
[total] 5177.0 1872.0
OnlineSecurity
No 2038.0 1462.0
Yes 1725.0 296.0
Nointernetservice 1414.0 114.0
[total] 5177.0 1872.0
OnlineBackup
Yes 1907.0 524.0
No 1856.0 1234.0
Nointernetservice 1414.0 114.0
[total] 5177.0 1872.0
DeviceProtection
No 1885.0 1212.0
Yes 1878.0 546.0
Nointernetservice 1414.0 114.0
[total] 5177.0 1872.0
TechSupport
No 2028.0 1447.0
Yes 1735.0 311.0
Nointernetservice 1414.0 114.0
[total] 5177.0 1872.0
StreamingTV
No 1869.0 943.0
Yes 1894.0 815.0
43

Nointernetservice 1414.0 114.0
[total] 5177.0 1872.0
StreamingMovies
No 1848.0 939.0
Yes 1915.0 819.0
Nointernetservice 1414.0 114.0
[total] 5177.0 1872.0
Contract
Month-to-month 2221.0 1656.0
One year 1308.0 167.0
Two year 1648.0 49.0
[total] 5177.0 1872.0
PaperlessBilling
Yes 2772.0 1401.0
No 2404.0 470.0
[total] 5176.0 1871.0
PaymentMethod
Electronic check 1295.0 1072.0
Mailed check 1305.0 309.0
Bank transfer (automatic) 1287.0 259.0
Credit card (automatic) 1291.0 233.0
[total] 5178.0 1873.0
MonthlyCharges
mean 61.2651 74.4415
std. dev. 31.09 24.6591
weight sum 5174 1869
44

precision 0.0634 0.0634
TotalCharges
mean 2555.3435 1531.7976
std. dev. 2329.237 1890.3342
weight sum 5163 1869
precision 1.3273 1.3273
Decision Table
45

Correctly Classified Instances 5567 79.043 %
Incorrectly Classified Instances 1476 20.957 %
Kappa statistic 0.4212
Mean absolute error 0.2911
Root mean squared error 0.3775
Relative absolute error 74.6522 %
Root relative squared error 85.501 %
Total Number of Instances 7043
8.2Results
In the web usage mining, there are two types of clustering, they are web transaction
clustering and internet page clustering that is executed at usage records . One successful utility of
internet web page clustering is the adaptive internet web site. As an instance, the set of rules
called PageGather5. To synthesize index pages, PageGather5 is proposed that don't exist to begin
with, based on partitioning web pages into various agencies.
The diverse get admission to pastimes of customers in step with their navigational
histories are used to represent the generated index pages. Some other example is that clustering
consumer rating results has been successfully adopted in collaborative filtering applications as a
facts making ready step to improve the scalability.
46

Recommended to the use of k-Nearest-Neighbour (kNN) algorithm:
When doing the analysis of data set through WEKA tool, k-Nearest Neighbour algorithm
is used. Because it classifies 5567 instances correctly rather than classification and clustering
algorithms.
The internet transaction and web page clustering techniques, which is employing the
traditional k-means clustering set of rules to signify person get entry to patterns for net
personalization primarily based on mining web usage facts. Those proposed clustering-based
strategies had been proven to be efficient from their experimental results for the reason that they
may be honestly able to figuring out the intrinsic common attributes found out from their
historical clickstream statistics. Normally, these utilization styles are explicitly captured at the
extent of user consultation or web page. They, however, do now not monitor the underlying
characteristics of user navigational sports in addition to web pages. As an example, such located
utilization patterns provide little statistics of why such net transactions or group of web pages,
and latent relationships most of the co prevalence observation information have now not been
incorporated into the mining procedures as properly.
For this reason, its miles essential to develop LSA-based tactics that could screen not
most effective not unusual tendencies explicitly, but additionally take the latent statistics into
account implicitly all through mining. An set of rules primarily based on most important factor
evaluation (PFA) version derived from statistical evaluation, is proposed to generate consumer
get right of entry to patterns and uncover latent factors by clustering consumer transactions and
reading foremost elements involved inside the internet usage mining. Analogous, some research
are addressed to derive user get admission to patterns and internet page segments from diverse
kinds of internet statistics, by way of using a so-called Probabilistic Semantic Latent analysis
(PLSA) version, that is based totally on a most likelihood precept from records.
Comparison of E-commerce website usage
The benefit of online retailers is to get pricing as they don’t spend expenses like save
rent, payments etc. Compare than offline market, the online retailers pass by their rate
immediately to customer and generally offer a lower charge. The offline shopping is better than
online shopping even they applied some delivery charges. Hence, determines the level of online
shopping. Lower the price- higher the temper to demand, higher the price –lower the demand.
47

Price of any commodity also influences the purchasing electricity of any precise buyer.
Consumer normally pick mild or everyday charge with excellent quality and do not favor to
spend or fee lot for any form of stuff. So, shopping for and promoting each are simultaneously
affected by the fee of product. To achieve this, various categorized shopping products are
available in the developed E-Commerce website such as books, clothes, music and so on (Gupta,
2015). The below screenshot shows the result of the categorized products.
The analysis regarding user behavior towards online shopping is given below (Gupta.
2015).
48

The above graphical illustration indicates the age class of populace who choose online
shopping. The clients who use online shopping are mostly between 18 to 25 age purchase 68.6%
of the total buyer because of the growing technological revolution and the youth population
compare than the different age groups. Next, the age group of 25-35 are the second most
consumer type using online purchasing services of 14.7%. The main reason for this age group is
time. The rest of the buyers are the age above than 35 and below 18 purchase 8.7% and 8.0%
respectively. The percentage of populace low because most of the 35 above group have lack of
ample understanding of technological know-how used. And for 18 below cash constraint comes
into picture.
The experimental analysis was done for the questionnaire, “Is website providing the
sufficient Information?” The below chart shows that the result for the survey data.
The web page offers well-informed statistics for every product along with images to
entice the clients. The online website provide the valuable information needed for customer and
many instances its prevented when the product is not up to the mark in 69% .every information
needed is supplied which makes the product administered person pleasant and a purchaser can
purchase the product judiciously. Only in 5% cases it was once viewed that the product data is no
longer supplied and 26 percent of the respondent said that it grant the records sometime.
49

The below given chart displays the survey of the questionnaire, “Does online shopping is
as secure as traditional shopping?”
The Graph shows what customers assume about online shopping, with 42.70% sometime
suppose it’s impervious someday it is not as in contrast to typical shopping.40% consumer
suppose online purchasing is as secure as common shopping, while 17.30 % accept as true with it
is no longer secure.
To implement the web application development for an efficient E-Commerce website,
Internet Beans IDE is used. Java is used for front end the SQL is used for backend. The user,
administrator and employees can login and access the functions as given such as viewing the
products, searching, shopping and so on. The list of available products and discounts on those
products as shown below.
The proposed system facilitates that the user can review the products for easy buy
according to their financial capability and favourites. This session performs the research while
asking the questions from the customer who are purchasing by offline and usual to charge each
the channels in the terms of performance, products, time of delivery, high-quality and other
associated components of on-line and offline shopping. Investigators and professionals can
obtain treasured insights to undertake a precise channel for shopping. A cluster of all customer
communication with products, offerings or people that is specific and the conceptual model, a
theoretical context for the find out about of the theoretical groundwork of all the research on it is
placed. Customer behavior is a method of discussions with investors to purchase patron items
and purchaser behavior guidelines of the manner (Mittal, 2013).
50

The following steps are followed by the ordinary commercial enterprise customer: To
identify a need or problem: inside the change press, the press insurance or advertising and
marketing are visible. Growing product specification: The client needs are fulfilled by some
sources that can locate to identify their needs. They’ll pay specific interest to press releases,
exhibitions, editorial comment, advertising, industry seminars and applicable direct mail.
business client is especially open to visits out of your sales force and alternate directory entries.
Exhibitions and technical facts leaflets are also useful assets. That is the time whilst pricing
statistics starts off evolved to be severely taken into consideration.
8.3Summary
The result and discussion is about simply identify techniques called web Mining,
architectures and tactics is the fastest of utilising records mining techniques to extract and induce
useful facts information from internet, services and online goods will increase, information
mining sports can expand companies to rapidly allowing to retrieve exceptionally personalized
statistics approximately customers, which as nicely implies concerns and violations which are
highly privacy. Each marketers and customers have to follow privacy coverage policies.
Entrepreneurs ought to pay more interest to degree of user consider and couple their records
mining performance with respect to person privacies. In this dissertation, the research gives a
business and technological assessment of records mining and outlines how, together with sound
commercial enterprise approaches and complementary technologies, consumer relationships can
be redefined and reinforced by the data mining technique.
9. Identifying Users’ Behavioural Patterns
User Behavioural Analysis
From the study of various user behavioural analysis towards E-shopping websites, the
time to display the sorted out products should be very small which means the query processing
should not take much time. To achieve this aim, an efficient searching algorithms will be applied
for product sorting. Binary search algorithm is considered as the best one to produce result in
best case. So the analysis will be done on these algorithms used on various websites. Further to
knowledge the desires of the customers, enterprise also want to apprehend what motivates them
to buy, and the way can affect the shopping for process to make sure that the products or services
51

are on the shopping listing. Understanding the customers will assist, to increase and distribute the
product, in addition to getting the proper price point and developing a hit promotional sports.
The psychology of the buying manner has been broadly studied and irrespective of what length
company business, information of this process can assist organisation grow to be greater
successful. Each businesses and clients show off patterns of buying behaviour. The commercial
enterprise model is much less open to discuss as the enterprise clients will almost actually have
some formalized manner of buying in place. The corporation undertaking is to understand the
procedure and in shape the marketing sports to the specific tiers of the system. Because of this
the Patron will get hold of the right kind of contact on the proper time. To provide insights into
regions such as price sensitivity, segmentation, and client wishes evaluation, indicators of
consumer defection.
9.1Summary
On this patterns, In our internet recommended framework, we use the model-primarily
based technique. We do not forget the used consumer profiles that is generated in preceding
segment because the common navigational behaviours that is represented by the aggregated
manner exhibited by all individuals in the identical unique person category, and make use of
them as a usage information base for recommending probably visited web pages to the present
day person. Just like the approach proposed for representing consumer get admission to use in
the shape of weighted page vector that has an n-dimension, we make use of the cosine function
to degree the similarity among the current lively person consultation and observed usage
patterns. We, then, pick the nice suitable profile, which stocks the best similarity with the
cutting-edge consultation, as the matched sample of current person.
10.Conclusions and Future Work
The project in particular centred on the analysis of client behaviour closer to e-commerce
internet site. At present, the humans around the globe are addicted to on-line buying because of
its monstrous consolation for the customers. E-commerce has come to be so commonplace and
famous because, it brings ease into human beings’ lives. These days’ E-commerce websites are
extraordinarily effective and powerful to gain the clients. The E-commerce websites provide n
52

range of types at one location. It is understood that e-commerce specifically makes a speciality of
customer pleasure. Preserving all of the benefits of E-commerce in mind, it is vital to know that,
it is complex to understand the behaviour of every client, which helps in enjoyable the clients. A
session of the person on different web sites is stored, which allows in figuring out the
behavioural styles of the customers. But, it's far viable to seize the records like, what the person/
client would like to buy, based totally on more often than not regarded merchandise. The web
server shops this type of data within the form of web logs. The logs helps to investigate the
information to realize the behaviour of the users. Consequently, this venture works on expertise
the clients and their interested products, to satisfy them. To understand the fundamental of e-
commerce internet site, a deep have a look at has been performed. On this component includes
how e-commerce emerged, why human beings decide upon on line buying over offline
purchasing, and the way it works in existing machine.
To investigate the patron behaviour toward e-trade website, 3 model checking is finished.
The models are Linear Temporal logic version, Computational temporal logical version and
Probabilistic version. The stability of every node’s nation is evaluated. An e-commerce website
is advanced to investigate the client behaviour primarily based at the parameters assessment of
merchandise and providers, purchaser shopping for conduct, the customer's want, finding out
opportunity merchandise and providers and buy decision. The obtained consequences are
mentioned in element. To expand e-commerce website, Java is used. To make the internet site
for consumer pleasant, the Bootstrap is carried out. The sample Behavioural patterns are taken
and those are carried out on Weka tool for statistics pre-processing. The evaluation component is
completed with evaluating 3 pinnacle maximum e-trade website and how the people are
appearance and feel whilst on-line buying. Charts and tables are used in this analysis element for
clean expertise of information. For that reason, powerful assessment is the key to finish the
evaluation of this task. For accomplishing the evaluation the help of net analysts is taken to
concentrate on making use of records mining techniques inclusive of k means nearest neighbour
(KNN) algorithm for studying person’s conduct. Sequence of actions is completed. Additionally,
the behavioural patterns help to understand the interests of users, to be applied on E-Commerce
web sites. The linear-temporal common sense model is proposed in this task to assist the
evaluation of the structured e-commerce web logs. Based on the E-trade structure, a common
technique for mapping the log facts can assist in converting them to occasion logs, which
53

captures the customers’ behaviours or moves. Later, diverse predefined queries are utilized for
identifying the behavioural patterns which considers various actions achieved through the user,
at some point of a session. As a very last factor, the proposed model is implemented on actual
case studies of different E-commerce websites. On the other hand, within the software program
development life cycle, the waterfall model is used, which looks after requirement amassing,
planning, designing, coding and checking out. The consequences of the analysis are provided.
References
Bejju, A. (2018). Sales Analysis of E-Commerce Websites using Data Mining
Techniques. International Journal of Computer Applications (0975 – 8887), 133(5).
De Giacomo, G., De Masellis, R. and Montali, M. (2014). Reasoning on LTL on Finite Traces:
Insensitivity to Infiniteness. Association for the Advancement of Artificial Intelligence.
Gupta, P. (2015). Comparative Study of Online and Offline Shopping: A Case Study of Rourkela
in Odisha. Department of Humanities and Social Sciences National Institute of Technology
Rourkela. MA Thesis. Available at: [Accessed 8 Mar. 2018].
Kim, Y. and Yum, B. (2011). Recommender system based on click stream data using association
rule mining. Expert Systems with Applications, 38(10), pp.13320-13327.
Mittal, A. (2013). E-commerce: It’s Impact on consumer Behavior. Global Journal of
Management and Business Studies, 3(2), pp.131-138.
Moe, W. and Fader, P. (2004). Dynamic Conversion Behavior at E-Commerce
Sites. Management Science, 50(3), pp.326-335.
Orendorff, A. (2017). Global Ecommerce Statistics [Infographic] and 10 International Growth
Trends You Need to Know. [online] Enterprise Ecommerce. Available at:
https://www.shopify.com/enterprise/global-ecommerce-statistics [Accessed 8 Mar. 2018].
Poggi, N., Muthusamy, V., Carrera, D. and Khalaf, R. (2013). Business Process Mining from E-
Commerce Web Logs. Lecture Notes in Computer Science, pp.65-80.
54

Sergio, H., Alvarez, P., Fabra, J. and Ezpeleta, J. (2018). Analysis of users’ behaviour in
structured e-commerce websites. ACCEPTED FOR ITS PUBLICATION IN IEEE ACCESS.
Vasconcelos, L., Santos, R. and Baldochi, L. (2016). Exploiting Client Logs to Support the
Construction of Adaptive e-Commerce Applications. IEEE, 13th International Conference on e-
Business Engineering (ICEBE), pp.164-169.
Appendix
To accomplish the secured website, in this project, the jdbc:odbc connection is established. The
jdbc:odbc connection code is given below.
importjava.sql.Connection;
importjava.sql.DatabaseMetaData;
importjava.sql.DriverManager;
importjava.sql.SQLException;
publicclassJdbcSQLServerConnection {
publicstaticvoidmain(String[] args) {
Connection conn = null;
try{
String databaseURL = "jdbc:DESKTOP-7G6UMKK://localhost\\
sqlexpress";
String user = "root";
String pass = "root";
conn = DriverManager.getConnection(databaseURL, username,
password);
if(conn != null) {
DatabaseMetaData dmd = (DatabaseMetaData) conn.getMetaData();
System.out.println("Driver name: "+ dmd.getDriverName());
System.out.println("Driver version: "+
dmd.getDriverVersion());
System.out.println("Product name: "+
dmd.getDatabaseProductName());
System.out.println("Product version: "+
dmd.getDatabaseProductVersion());
}
} catch(SQLException ex) {
ex.printStackTrace();
} finally{
try{
if(conn != null&& !conn.isClosed()) {
55

conn.close();
}
} catch(SQLException ex) {
ex.printStackTrace();
}
}
}
}
The below figure shows that the database connectivity done in this web application
development.
56