A MINI PROJECT REPORT ON

BOOK RECOMMENDATION SYSTEM

Submitted in partial fulfillment of the requirements for the award of Bachelor of

Technology in Information Technology

By

ARGYADIP DAS (21911A1206)

A.SUBRAMANYAM GUPTA ( 21911A1205)

PRANEETH TEJA (21911A1223)

Under the Esteemed Guidance of

Mr.Suresh Babu

Assoc.Professor

DEPARTMENT OF INFORMATION TECHNOLOGY

VIDYA JYOTHI INSTITUTE OF TECHNOLOGY
(Accredited by NBA, Approved by AICTE, Affiliated to JNTU
Hyderabad) Aziz Nagar Gate, C.B.Post, Chilkur Road,
Hyderabad – 500075 2021 - 2025
 DEPARTMENT OF INFORMATION TECHNOLOGY

CERTIFICATE

This is to certify that the Project Report on “Book Recommendation System” is a bonafide
work by Argyadip Das (21911A1206), Subramanyam (21911A1205),Praneeth(21911A1223)
in partial fulfilment of the requirement for the award of the degree of Bachelor of Technology
in “INFORMATION TECHNOLOGY” JNTU Hyderabad during the year 2024-2025.

Project Guide
Mr.B.Suresh Babu, M.Tech,Associate Professor

Head of the Department

Mr.B.Srinivasulu, M.E,Professor

Internal Examiner External Examiner

ii
 DECLARATION

We, Argyadip Das(21911A1206), A.Subramanyam (21911A1205),

Praneeth Teja(21911A1223) here by declare that the project entitled
“BOOK RECOMMENDATION SYSTEM” done by us under the
guidance of Mr.B.Suresh Babu ,M.Tech,Assoc.Professor is submitted in
partial fulfillment of the requirements for the award of Bachelor of
Technology degree in Infromation Technolgogy to Vidya Jyothi Institute
of Technology , affiliated to JNTU-Hyderabad, is an authentic work and
has bot been submitted to any other university or institute for the
degree.

Argyadip Das(21911A1206)
A.Subramanyam(21911A1205)
Praneeth Teja(21911A1223)

iii
 ACKNOWLEDGEMENT

I am pleased to acknowledge my sincere thanks to Board of Management of

VIDYA JYOTHI for their kind encouragement in doing this project and for
completingit successfully. Iam grateful to them.

I convey my thanks Mr.B.Srinivasulu M.E., Professor, Head of the Department of

Information Technology for providing me necessary support and details at the right
time duringthe progressive reviews.

I would like to express my sincere and deep sense of gratitude to my Project Guide
Mr.Suresh Babu,Associate Professor, for his valuable guidance, suggestions and
constant encouragement paved the way for the successful completion of our project
work.

I wish to express my thanks to all Teaching and Non-teaching staff members of the
Department of Information Tehnology who were helpful in many ways for the
completion of the project.

iv
 TABLE OF CONTENTS

Chapter
TITLE Page No.
No.
ABSTRACT v

LIST OF FIGURES vii

1 INTRODUCTION 1

1.1 Problem Definition 2

1.2 Scope and Objectives 2

2 LITERATURE SURVEY 3

2.1 Inferences from Literature Survey 5

2.2 Motivation 6

2.3 Open problems in Existing System 7

3 REQUIREMENTS ANALYSIS 8
3.1 Feasibility Studies / Risk Analysis of the Project 8
3.2 Software Requirements Specification Document 8
4 DESCRIPTION OF PROPOSED SYSTEM 13
4.1 Selected Methodology or process model 13
4.2 Data sets 14

4.3 Architecture / Overall Design of Proposed System 18

Description of Software for Implementation and Testing 19
4.4
plan of the Proposed Model/System
4.5 Project Management Plan 24
5 IMPLEMENTATION DETAILS 32

5.1 Development and Deployment Setup 25

5.2 Algorithms 26
5.3 Testing 26
6 RESULTS AND DISCUSSIONS 32
7 CONCLUSION 33
7.1 Conclusion 33
v
7.2 Future Work 33
7.3 Research Issues 34

REFERENCES 35
APPENDIX
A. SOURCE CODE
B. SCREENSHOTS
C. RESEARCH PAPER

vi
 LIST OF FIGURES

FIGURE FIGURE NAME PAGE

NO. NO.
2.1 Basics Statistics 4
2.2 Global Recommendation System 6
4.1 Collaborative Filtering Process 17
4.2 Pearson Correlation 18
4.3 System Architecture 19
4.4 Localhost web Server 22
4.5 Localhost Output Console 23
4.6 Flowchart 24

vii
 ABTRACT

Today the World Wide Web provides users with a vast array of information, and
commercial activity on the Web has increased to the point where hundreds of new
companies are adding web pages daily. This has led to the problem of information
overload. Recommender systems have been developed to overcome this problem
by providing recommendations that help individual users identify content of interest
by using the opinions of a community of users and/or the user’s preferences.

The aim of this thesis was to design and evaluate different approaches for producing
personalised recommendations within the book domain. To achieve this goal, the
project first investigated existing recommender systems and profiling techniques.
The next step was to build users’ profiles by monitoring users’ behaviour, and
develop three different approaches for producing recommendations. Finally, an
evaluation of the system recommendations’ accuracy was done, by first conducting
live user experiments and then performing offline analysis to measure the
recommendations’ accuracy using appropriate methods for testing.

The system evaluation results show that the accuracy of the system
recommendations is very good and that a recommender system based on the
combination of content-based and collaborative filtering approaches provides more
accurate recommendations for the book domain.

viii
 CHAPTER 1
INTRODUCTION

This is an experimental project which first, designs…. And second evaluates

different approaches for offering recommendations to readers regarding books they
may wish to purchase, as part of an online bookshop website.

Today the World Wide Web has provided access to a vast array of information
through the web pages, as a result of the Internet growth. Also, commercial activity
on the Web has increased to the point where hundreds of new companies are adding
Web pages daily. With this increase in the information sources, a problem of
information overload occurs, in which the users are trying to deal with an excess of
information that is not useful to them as they try to make sensible decisions (Losee,
1989). As a response to this problem, a range of tools to help with retrieving,
searching, and filtering have been developed.

The tool most widely used to alleviate the problem of information overload is the
search engine. The benefits for the users from search engine technology have
decreased as the number of web pages has grown. In addition, the user must first
consider the large number of search tools available and decide which one to access.
Then the user must interact with each one individually because search engines are
typically not personalised to individual users or their prevailing context. Users usually
make a choice on the basis of their personal experience or other people’s
experience. Based on these facts, recommender systems have been developed to
provide recommendations that help individual users identify content of interest by
using the opinions of a community of users and/or the user’s preferences.

Today various recommendation systems play an important role in supporting

commercial websites to help users find items that they know they would like to
purchase, as well as discover new items about which they had been unaware. The
ability to persuade the consumers to buy a suitable item is a significant goal for any
recommender system in an ecommerce environment. However, for any
recommender system to be successful, the consumer must trust and accept the

1
system’s recommendations. This is done with a clear explanation from the system,
presented in a way that is in keeping with the consumer’s preferences. A good
recommender system can significantly contribute to achieving the consumer’s
acceptance of the system recommendations.

1.1 PROBLEM DEFINITION

This project aims to design and evaluate different approaches for computing
recommendations within the book domain to provide personalised
recommendations to the users.

1.2 SCOPE AND OBJECTIVES

• An effective solution to the issue of information overload in e-commerce
websites is the recommender system.
• This method offers users accurate recommendations.
• most reliable book-related suggestion technology.

Objectives:
• Look into and assess the profiling and recommender systems that are
already in use.
• By observing dynamic user behaviours, you can create a user's profile
for a recommender system. The user profile needs to change to reflect
the user's shifting interests.
• Create a recommender system that uses a variety of computation
methods.
• Utilize the right methods to assess the system's recommendations'
accuracy.

2
 CHAPTER 2
LITERATURE SURVEY

An idea for Content-Based Book Recommending Using ML Tools for Text

Categorization was put forth by Raymond J. Mooney and Loriene Roy. They outline
a content-based book recommendation system that classifies text using machine
learning and information extraction. Item recommendations are made based on
information about the item itself rather than on the preferences of other users. On
the other hand, learning customiZed profiles from descriptions of examples enables
a system to uniquely characteriZe each customer without the need to match his or
her interests to another's. Through the use of automatic text-categorization
techniques on semi-structured text downloaded from the web, they have been
investigating content-based book recommendations. A database of book data is
used by the present prototype system, called LIBRA (Learning Intelligent Book
Recommending Agent). After employing a Bayesian learning algorithm to learn
about the user's profile, the system generates a ranked list of the most highly
suggested additional books from its library. Even when the method provides very
modest training sets, the overall results are quite positive. An original content-based
book recommender called LIBRA makes use of a straightforward.

A unique book recommendation system was proposed by Binge Cui and Xin Chen.
When readers are unable to locate the desired book using the library's bibliographic
retrieval system, they are directed to the recommendation pages. It is a web-based
system for recommending books to a library's patrons. After logging in, a user can
search for books using author names or keywords like book titles. A bibliographic
retrieval system will then look for books using the same keywords. If the
recommendation system returns any results, submit these keywords to the web
books retrieval module. Web Books Retrieval Module allows the librarian or
administrator of the online book recommendation system to search the online
bookshop using keywords by creating accounts on sites like Amazon. As a result,
the web retrieval module searches these online bookshops as the logged-in user
when the keyword is presented to it. The user will receive the results from these
online booksellers in the form of recommendations. The statistic and analysis

3
module will determine the value of that specific book based on user
recommendations. The Auto-Order Module will then generate a book order
automatically based on the analysis results according to this value of book. The
Short Message and Email Notification Module will get a report from the Book Storage
System once the purchased books have been shelved (fig 2.1). Then, utilizing a
message and email server, it will inform the readers who have suggested the books
that have been acquired.

Fig 2.1 Basic Statistics

An effective and best-in-class hybrid recommendation algorithm was proposed by

Dharmendra Pathak, Sandeep Matharia, and C. N. S. Murthy , and it provides
recommendations that are more in line with user preferences. The hybrid
recommendation in this case combines collaborative, content-based, and context-
based algorithms. The primary input for collaborative filtering is rating, or the votes
of numerous users, content-based data, which is user-specific information like
interests, dates of birth, and priorities, and context-based data, which is behavioural
information like time, taste, mood, and weather. The similarity is measured using the
cosine similarity. According to the user's prior history, there are subject priority.
When they buy a book, do they check to see if the subject priority is changed from
what was previously set? If so, topic priority 3 and subject priority 2 should be reset.
Priority 1 will remain the same. They came to the conclusion that the proposed
Hybrid book recommendation algorithm is superior to the others based on

4
calculations and results.

A book recommendation system based on integrated features of content filtering,

collaborative filtering, and association rule mining was proposed by Anand Shanker
Tewari, Abhay Kumar, and Asim Gopal Barman. When a customer searches for a
book, the search is recorded in the customer's purchase or search history.
Recommendation services conduct some filtering when a customer is not online,
and the results are saved in the customer's web profile. The recommendations will
be created automatically the following time the customer visits the website. Web
Usage Mining (WUM) is used in content-based filtering to give customers the
pertinent information they need. web server access logs, browser caches, or proxy
logs are some examples of historical data that web Usage Mining (WUM) commonly
uses to extract knowledge. web use Internet user behavior is recorded by mining,
which then processes the information (fig 2.2). The item-based collaborative
recommendation algorithm and cosine similarity are both used to measure similarity.
Compare the outcomes of collaborative content filtering with association rule mining.

Fig 2.2 Global recommendation system

2.1 INFERENCES FROM LITREATURE SURVEY

For a book recommendation system, Adli Ihsan Hariadi and Dade Nurjanah
presented a hybrid-based approach that blends attribute-based and user
personality-based methodologies. The MSV-MSL (Most Similar Visited Material to
the Most Similar Learner) method is used in this study, and the authors claim that it
is the best hybrid attributes-based strategy. When forming neighbourhood ties, the
personality trait is used to compare users. The hybrid attribute will use the similarity

5
scores between a target book and its neighbours as well as between the active user
and that user's neighbours to generate the recommendation scores of rated books
from neighbours. the score for user u's book B, designated as score b. The goal of
this is to match up the most similar learner with the most similar visited material. It
makes advantage of the collaborative and content values. Utilize the hybrid result
as a recommendation after that. That is the webpage that the most similar learner
has visited, according to data.

2.2 MOTIVATION

Due to the expansion of the Internet, the World Wide Web now offers access to a
wide variety of information via web sites. Additionally, business activity online has
grown to the point that every day, hundreds of new businesses add new Web pages.
Due to the increase in information sources, a problem known as information
overload arises, where users must cope with an abundance of information that is
unhelpful to them in order to make sound judgments (Losee, 1989). A variety of tools
for accessing, finding, and filtering information have been created as a solution to
this issue The search engine is the resource that is most frequently utilised to
address the issue of information overload. As the amount of web pages has
increased, so too have the benefits for consumers of search engine technology.
Additionally, the user must choose which search tool to utilise after carefully
weighing the several options. Then, as search engines are often not tailored to
specific individuals or their current environment, the user must interact with each
one separately. Users typically base their decisions on either their own or other
people's experiences. These facts led to the development of recommender systems,
which leverage user feedback from a community of users and/or the user's own
preferences to generate recommendations that assist individual users in identifying
content of interest.

2.3 OPEN PROBLEMS IN EXISTING SYSTEM

The content of the Web is expanding by an estimated 170,000 pages every
day, with an Internet growth rate of at least 10% per month (Turban et al., 2000).
Due to the abundance of information available to consumers on the World Wide
Web, it can be challenging for them to pick exactly what they want. Information

6
overload is a condition when consumers strive to manage more information but are
unable to make rational decisions (Losee, 1989).
A variety of retrieval, searching, and filtering techniques have been created
to aid with the problem of information overload. The search engine is the most
frequently used instrument to help with the issue of information overload. Although
search engines are efficient in filtering pages, consumers find it challenging and
time-consuming to express their needs in a search query. Due to the exponential
growth of web pages, search engine technology's advantages for users have
lessened with time.

7
 CHAPTER 3

REQUIREMENT ANALYSIS

3.1 FEASIBILITY STUDIES / RISK ANALYSIS OF THE PROJECT

The project will create and assess a collaborative filtering and content-based
recommender system for a real online bookstore. Machine learning methods are
typically needed for content-based recommendations in order to identify trends in
the products customers like (Middleton, 2003). The experiences of actual users will
be reflected in the content-based technology. Users' profiles will be created so that
their behaviour may be tracked. Additionally, the system will produce
recommendations by comparing the contents of the books in the user's profile with
those that the user hasn't reviewed.

3.2 SOFTWARE REQUIREMENTS SPECIFICATION DOCUMENT

3.2.1 Hardware Requirements

• System : Intel® Core™ i5-9300H CPU @ 2.40GHz.

• Monitor : LED.

• Mouse : Logitech.

• Ram : 8.00 GB or above 8.00 GB

• Hard Disk : 1 TB

3.2.2 Software Requirements:

• Operating System : Windows 10, Kali Linux

• Language : Python 3

• Framework : Flask 2.2.2

3.2.3 Python:
Python is a high-level, interpreted, interactive and object-oriented scripting

8
language. Python is designed to be highly readable. It uses English keywords
frequently where as other languages use punctuation, and it has fewer syntactical
constructions than other languages.
• Python is Interpreted − Python is processed at runtime by the interpreter.
You do not need to compile your program before executing it. This is similar
to PERL and PHP.

• Python is Interactive − You can actually sit at a Python prompt and interact
with the interpreter directly to write your programs.

• Python is Object-Oriented − Python supports Object-Oriented style or

technique of programming that encapsulates code within objects.

• Python is a Beginner's Language − Python is a great language for the

beginner-level programmers and supports the development of a wide range
of applications from simple text processing to WWW browsers to games.

3.2.4 History of Python

Python was developed by Guido van Rossum in the late eighties and early nineties
at the National Research Institute for Mathematics and Computer Science in the
Netherlands.
Python is derived from many other languages, including ABC, Modula-3, C, C++,
Algol-68, Small-Talk, and Unix shell and other scripting languages.
Python is copyrighted. Like Perl, Python source code is now available under the
GNU General Public License (GPL).
Python is now maintained by a core development team at the institute, although
Guido van Rossum still holds a vital role in directing its progress.
3.2.5 Python Features
Python's features include −
• Easy-to-learn − Python has few keywords, simple structure, and a clearly
defined syntax. This allows the student to pick up the language quickly.

• Easy-to-read − Python code is more clearly defined and visible to the eyes.

• Easy-to-maintain − Python's source code is fairly easy-to-maintain.

9
 • A broad standard library − Python's bulk of the library is very portable and
cross-platform compatible on UNIX, Windows, and Macintosh.

• Interactive Mode − Python has support for an interactive mode which allows
interactive testing and debugging of snippets of code.

• Portable − Python can run on a wide variety of hardware platforms and has
the same interface on all platforms.

• Extendable − You can add low-level modules to the Python interpreter.

These modules enable programmers to add to or customize their tools to be
more efficient.

• Databases − Python provides interfaces to all major commercial databases.

• GUI Programming − Python supports GUI applications that can be created

and ported to many system calls, libraries and windows systems, such as
Windows MFC, Macintosh, and the X Window system of Unix.

• Scalable − Python provides a better structure and support for large programs
than shell scripting.

Apart from the above-mentioned features, Python has a big list of good features,
few are listed below −
• It supports functional and structured programming methods as well as OOP.

• It can be used as a scripting language or can be compiled to byte-code for

building large applications.

• It provides very high-level dynamic data types and supports dynamic type
checking.

• It supports automatic garbage collection.

• It can be easily integrated with C, C++, COM, ActiveX, CORBA, and Java.

Python is available on a wide variety of platforms including Linux and Mac OS

X. Let's understand how to set up our Python environment.
3.2.6 Getting Python
The most up-to-date and current source code, binaries, documentation, news, etc.,

10
is available on the official website of Python https://www.python.org.

3.2.7 Windows Installation

Here are the steps to install Python on Windows machine.
• Open a Web browser and go to https://www.python.org/downloads/.

• Follow the link for the Windows installer python-XYZ.msifile where XYZ is the
version you need to install.

• To use this installer python-XYZ.msi, the Windows system must support

Microsoft Installer 2.0. Save the installer file to your local machine and then
run it to find out if your machine supports MSI.

• Run the downloaded file. This brings up the Python install wizard, which is
really easy to use. Just accept the default settings, wait until the install is
finished, and you are done.

The Python language has many similarities to Perl, C, and Java. However, there
are some definite differences between the languages.

3.2.8First Python Program

Let us execute programs in different modes of programming.

Interactive Mode Programming

Invoking the interpreter without passing a script file as a parameter brings up the
following prompt −

$ python

Python 2.4.3(#1, Nov112010,13:34:43)

[GCC 4.1.220080704(RedHat4.1.2-48)] on linux2

Type "help", "copyright", "credits" or "license" for more information.

>>>

Type the following text at the Python prompt and press the Enter −

11
>>>print "Hello, Python!"

If you are running new version of Python, then you would need to use print
statement with parenthesis as in print ("Hello, Python!"). However, in Python
version 2.4.3, this produces the following result −

Hello, Python!

3.2.9 Script Mode Programming

Invoking the interpreter with a script parameter begins execution of the script and
continues until the script is finished. When the script is finished, the interpreter is no
longer active.
Let us write a simple Python program in a script. Python files have extension .py.
Type the following source code in a test.py file −

Print "Hello, Python!"

We assume that you have Python interpreter set in PATH variable. Now, try to run
this program as follows −

$ python test.py

This produces the following result −

Hello, Python!

12
 CHAPTER 4

DESCRIPTION OF PROPOSED SYSTEM

The application will be developed using the incremental development methodology

and will be made up of four increments: Front End, Learning module,
Recommendation module and Database increment. The requirements outlined in
the Requirements Document will be mapped to manageable increments.

4.1 SELECTED METHODOLOGY OR PROCESS MODEL

4.1.1 Methodology

Recommender systems have been developed to overcome the above mentioned

limitations of searching through the massive volume of information available.
Recommender systems, in comparison with other filtering tools, require less
experience on the part of the user and less effort to specify their interests when
querying and operating the system (Resnick and Varian, 1997).
Recommendations systems rely on different technologies for computing
recommendations. The most important approaches are content-based filtering and
collaborative filtering. Content-based filtering displays users as individuals, while
recommender systems employing the collaborative filtering approach display the
user as a part of a group (Fasli, 2006). In addition, an advanced recommender
system that combines content-based and collaborative filtering to avoid the
limitations of each approach, is called a hybrid approach.

4.1.2 MODULE DESCRIPTION

Content based filtering approach
The content-based filtering approach identifies the similarity between a user and the
new items using the content of the previously evaluated items in the user profile. In
addition, each item in a user profile is characterized by a set of attributes which is
constructed by extracting a set of features from an item. Such a profile is used to
determine if the new item is similar to the item that a user has preferred in the past.
For instance, the Newsweeder is a netnewsfiltering system that suggests news
articles to the user based on the user’s profile (Lang, 1995). Most content-based

13
approaches are performed on textual documents, such as web pages and articles.
The textual document can be easily broken down into individual words, unlike video
and physical resources, which required sophisticated analysis.
Collaborative filtering approach
Collaborative filtering recommendations are based on the opinions of a community
of similar users. The basic idea is that users recommend items to one another.
Collaborative filtering makes this possible by asking the users to rate items, which
allows the system to recommend new items that similar users have rated highly. For
instance, MovieLens is a movie recommender system that uses collaborative
filtering to help people find movies they will like in the huge stream of available
movies. Collaborative filtering works well for multimedia technology such as music
and movies.
Data Set
During the last few decades, with the rise of Youtube, Amazon, Netflix and many
other such web services, recommender systems have taken more and more place
in our lives. From e-commerce (suggest to buyers articles that could interest them)
to online advertisement (suggest to users the right contents, matching their
preferences), recommender systems are today unavoidable in our daily online
journeys.

In a very general way, recommender systems are algorithms aimed at suggesting

relevant items to users (items being movies to watch, text to read, products to buy
or anything else depending on industries).

Recommender systems are really critical in some industries as they can generate a
huge amount of income when they are efficient or also be a way to stand out
significantly from competitors. As a proof of the importance of recommender
systems, we can mention that, a few years ago, Netflix organised a challenges (the
“Netflix prize”) where the goal was to produce a recommender system that performs
better than its own algorithm with a prize of 1 million dollars to win

4.2 DATA SET

Books are identified by their respective ISBN. Invalid ISBNs have already been

14
removed from the dataset. Moreover, some content-based information is given
(Book-Title, Book-Author, Year-Of-Publication, Publisher), obtained from Amazon
Web Services. Note that in case of several authors, only the first is provided. URLs
linking to cover images are also given, appearing in three different flavours (Image-
URL-S, Image-URL-M, Image-URL-L), i.e., small, medium, large. These URLs point
to the Amazon web site.
4.2.1 Recommender System
Recommender systems intend to provide users with suggestions of items that they
may be interested in, based upon their past preferences, history of purchase, or
demographic information, as well as the environment of possible items. In addition,
a recommender system helps the site adapt itself and provide individual
personalisation for each consumer; this increases the sales for the commercial site.

Different forms for providing recommendations have been developed; they can be
classified into the following forms: attribute-based recommendations, item-to-item
correlation, peopleto- people correlation and non-personalised recommendations
(Konstan et al., 2001). For more detailed descriptions.

Recommendations systems rely on different technologies for computing

recommendations. The most important approaches are content-based filtering and
collaborative filtering. Content-based filtering displays users as individuals, while
recommender systems employing the collaborative filtering approach display the
user as a part of a group (Fasli, 2006). In addition, an advanced recommender
system that combines content-based and collaborative filtering to avoid the
limitations of each approach, is called a hybrid approach.

4.2.2 Content based filtering approach

The content-based filtering approach identifies the similarity between a user and the
new items using the content of the previously evaluated items in the user profile. In
addition, each item in a user profile is characterized by a set of attributes which is
constructed by extracting a set of features from an item. Such a profile is used to
determine if the new item is similar to the item that a user has preferred in the past.
For instance, the Newsweeder is a netnewsfiltering system that suggests news
articles to the user based on the user’s profile (Lang, 1995). Most content-based

15
approaches are performed on textual documents, such as web pages and articles.
The textual document can be easily broken down into individual words, unlike video
and physical resources, which required sophisticated analysis.
Content-based filtering has some shortcomings in recommending items. A user's
selection is based on the subjective attributes (such as the quality) of the item
(Goldberg et al., 1992); in contrast, content based approaches are based on
objective attributes (such as the description of an item) about the items. Also, some
items the users may be interested in cannot be recommended to them because
content-based methods compare new items with the items previously seen by the
user, while the user's interests may be beyond the scope of the previously seen
items. Finally, multimedia technology such as sound, video or physical items cannot
be analysed automatically for relevant attribute information, due to limitations of
resources (Jennings et al., 2005).

4.2.3 Collaborative Filtering approach

Collaborative filtering recommendations are based on the opinions of a community
of similar users. The basic idea is that users recommend items to one another.
Collaborative filtering makes this possible by asking the users to rate items, which
allows the system to recommend new items that similar users have rated highly. For
instance, MovieLens is a movie recommender system that uses collaborative
filtering to help people find movies they will like in the huge stream of available
movies. Collaborative filtering works well for multimedia technology such as music
and movies. However, it also has some limitations:
New user problem: A new user starts off with a profile of interests from scratch.
The system needs to know the user preferences in different items to generate
accurate recommendations.
Cold start problem: New items cannot be recommended until more information is
obtained when another user either rates an item or provides feedback on the item
(Fasli, 2006). As a result, the recommendations generated by the system will not
recommend items similar enough to the users’ interests.
Scalability: A collaborative filtering algorithm should address the scalability issue
as the number of users increase and their collective profile size becomes large
(Fasli, 2006).

16
The schematic diagram of the collaborative filtering process is showed in Figure
4.1. As you can see from the figure, there is a list of users denoted by U= {u1,
u2,…,um} and a list of items I={i1,i2,….,in}. Each user has a list of items. The
collaborative filtering algorithm will generate recommendations(fig 4.1), a list of N
items that the active user will mostly like, according to the active user. Also, the
process will output a prediction, which is the result prediction on item j for the active
user (Sarwar et al., 2001).

Fig 4.1 Collaborative Filtering Process

4.2.4 Hybrid Approach

Hybrid approach introduced to combine the advantages of both content-based and
collaborative filtering techniques help to overcome their limitations. These use
thestrength of one set of filtering techniques to overcome the limitation of the other.
The hybrid filtering approach is also called “collaborative via content” because
content-based profiles are also taken when identifying the similarities among users
for collaborative recommendations (Pazzani, 1999).

4.2.5 User - Based collaborative Filtering

User-based algorithm is based on the fact that each user belongs to a larger group
of similarly behaving individuals. It uses statistical techniques to find a set of users
with similar interests, known as neighbours, in the entire user-item database, to
generate a list of recommendation for the active user (Middleton, 2003).
Different measures of similarity that are based on neighbourhood algorithms are
used to compute the similarity between the active user and other users in the
database, such as the Pearson correlation coefficient and Mean squared differences

17
algorithms (Breese et al., 1998). Moreover, to predict the rating of an item given by
the active user, the ratings from the most similar users for the item are averaged
and weighted by their similarities to the active user. The Pearson Correlation (fig
4.2) reflects the degree of linear relationship between two variables and ranges from
+1 to -1. A positive correlation means that the two users have very similar tastes,
while a negative correlation indicates that the users have dissimilar tastes (Fasli,
2006). The Pearson Correlation Coefficient method defines the similarity between
two users by:

Fig 4.2: Pearson correlation

4.2.6 Item Based collaborative Filtering

The item-based algorithms are developed to overcome the scalability on user-based
recommendations. Unlike a user-based approach, the item-based approach
identifies the set of items that are similar or related to the item that the active user
has evaluated. After that, it computes the similarity between items and then selects
the most similar items to the target item within the set of items that the user has
rated (Sarwar et al., 2001).

ALGORITHM
• Content based filtering mechanism
• Collaborative based filtering algorithm
• Cosine similarity

18
4.3 ARCHITECTURE / OVERALL DESIGN OF PROPOSED SYSTEM
Fig 4.3 depicts the architecture of the System

Fig 4.3: System Architecture

4.4 DESCRIPTION OF SOFTWARE FOR IMPLEMENTATION AND TESTING

PLAN OF THE PROPOSED MODEL/SYSTEM

Flask Framework:
Flask is a web application framework written in Python. Armin Ronacher, who
leads an international group of Python enthusiasts named Pocco, develops it. Flask
is based on Werkzeug WSGI toolkit and Jinja2 template engine. Both are Pocco
projects.
Http protocol is the foundation of data communication in world wide web. Different
methods of data retrieval from specified URL are defined in this protocol.

19
The following table 4.1 summarizes different http methods –

Table 4.1: Methods and description of the Flask

S.No Methods & Description

1 GET

Sends data in unencrypted form to the server. Most common method.

2 HEAD

Same as GET, but without response body

3 POST

Used to send HTML form data to server. Data received by POST method
is not cached by server.

4 PUT

Replaces all current representations of the target resource with the

uploaded content.

5 DELETE

Removes all current representations of the target resource given by a

URL

By default, the Flask route responds to the GET requests. However, this
preference can be altered by providing methods argument to route () decorator.

In order to demonstrate the use of POST method in URL routing, first let us create
an HTML form and use the POST method to send form data to a URL.

20
Save the following script as login.html

<html>

<body>

<form action=”http://localhost:5000/login” method="post">

<p>Enter Name:</p>

<p><input type="text" name="nm"/></p>

<p><input type="submit" value="submit"/></p>

</form>

</body>

</html>

Now enter the following script in Python shell.

from flask import Flask, redirect, url_for, request

app=Flask( name )

@app. route('/success/<name>')

def success(name):

return 'welcome %s'% name

@app. route ('/login’, methods=['POST','GET'])

def login ():

if request. method=='POST':

user=request. form['nm']

21
return redirect(url_for('success’, name= user))

else:

user=request.args.get('nm')

return redirect (url_for ('success’, name= user))

if name ==' main ':

app.run (debug =True)

After the development server starts running, open login.html in the browser, enter
name in the text field and click Submit.

Fig 4.4: Local Host Web Server

Form data is Posted to the URL in action clause of form tag.

http://localhost/login is mapped to the login () function. Since the server has

received data by POST method, value of ‘nm’ parameter obtained from the form
data is obtained by −

user = request. form['nm']

It is passed to ‘/success’ URL as variable part. The browser displays
a welcome message in the window.

22
 Fig 4.5: Local Host Output Console

Change the method parameter to ‘GET’ in login.html and open it again in the
browser. The data received on server is by the GET method. The value of ‘nm’
parameter is now obtained by −

User = request.args.get(‘nm’)
Here, args is dictionary object containing a list of pairs of form parameter and its
corresponding value. The value corresponding to ‘nm’ parameter is passed on to
‘/success’ URL as before.

23
4.5 PROJECT MANAGEMENT PLAN

Fig 4.6: Flow Chart

24
 CHAPTER 5

IMPLEMENTATION DETAILS

A recommendation engine is a class of machine learning which offers relevant

suggestions to the customer. Before the recommendation system, the major
tendency to buy was to take a suggestion from friends. But Now Google knows what
news you will read, Youtube knows what type of videos you will watch based on
your search history, watch history, or purchase history.

A recommendation system helps an organization to create loyal customers and build

trust by them desired products and services for which they came on your site. The
recommendation system today are so powerful that they can handle the new
customer too who has visited the site for the first time. They recommend the
products which are currently trending or highly rated and they can also recommend
the products which bring maximum profit to the company.

5.1 DEVELOPMENT AND DEPLOYMENT DETAILS

Machine learning is a process that is widely used for prediction. N number of
algorithms are available in various libraries which can be used for prediction. In this
article, we are going to build a prediction model on historical data using different
machine learning algorithms and classifiers, plot the results, and calculate the
accuracy of the model on the testing data.
Building/Training a model using various algorithms on a large dataset is one part of
the data. But using these models within the different applications is the second part
of deploying machine learning in the real world.
To put it to use in order to predict the new data, we have to deploy it over the
internet so that the outside world can use it. In this article, we will talk about how we
have trained a machine learning model and created a web application on it using-
Flask.
We have to install many required libraries which will be used in this model.
Use pip command to install all the libraries.

pip install pandas

pip install numpy

25
pip install sklearn

5.2 ALGORITHM
Content based Filtering: The algorithm recommends a product that is similar to
those which used as watched. In simple words, In this algorithm, we try to find finding
item look alike. For example, a person likes to watch Sachin Tendulkar shots, so he
may like watching Ricky Ponting shots too because the two videos have similar tags
and similar categories.
Collaborative based Filtering: Collaborative based filtering recommender systems
are based on past interactions of users and target items. In simple words here, we
try to search for the look-alike customers and offer products based on what his or
her lookalike has chosen. Let us understand with an example. X and Y are two
similar users and X user has watched A, B, and C movie. And Y user has watched
B, C, and D movie then we will recommend A movie to Y user and D movie to X
user.
Hybrid filtering method: It is basically a combination of both the above methods.
It is a too complex model which recommends product based on your history as well
based on similar users like you.

There are some organizations that use this method like Facebook which shows
news which is important for you and for others also in your network and the same is
used by Linkedin too.

Dataset description
we have 3 files in our dataset which is extracted from some books selling websites.

• Books – first are about books which contain all the information related to
books like an author, title, publication year, etc.
• Users – The second file contains registered user’s information like user id,
location.
• ratings – Ratings contain information like which user has given how much
rating to which book.

So based on all these three files we can build a powerful collaborative filtering
model. let’s get started.

26
Loading data
let us start while importing libraries and load datasets. while loading the file we have
some problems like.

• The values in the CSV file are separated by semicolons, not by a comma.
• There are some lines which not work like we cannot import it with pandas
and It throws an error because python is Interpreted language.
• Encoding of a file is in Latin

So, while loading data we have to handle these exceptions and after running the
below code you will get some warning and it will show which lines have an error that
we have skipped while loading.

Preprocessing Data: Now in the books file, we have some extra columns which
are not required for our task like image URLs. And we will rename the columns of
each file as the name of the column contains space, and uppercase letters so we
will correct as to make it easy to use.

The dataset is reliable and can consider as a large dataset. we have 271360 books
data and total registered users on the website are approximately 278000 and they
have given near about 11 lakh rating. hence we can say that the dataset we have is
nice and reliable.

We do not want to find a similarity between users or books. we want to do that If

there is user A who has read and liked x and y books, And user B has also liked this
two books and now user A has read and liked some z book which is not read by B
so we have to recommend z book to user B. This is what collaborative filtering is.

So this is achieved using Matrix Factorization, we will create one matrix where
columns will be users and indexes will be books and value will be rating. Like we
have to create a Pivot table.

If we take all the books and all the users for modeling, Don’t you think will it create
a problem? So what we have to do is we have to decrease the number of users and
books because we cannot consider a user who has only registered on the website
or has only read one or two books. On such a user, we cannot rely to recommend

27
books to others because we have to extract knowledge from data. So what we will
limit this number and we will take a user who has rated at least 200 books and also
we will limit books and we will take only those books which have received at least
50 ratings from a user.

Exploratory data analysis

The primary goal of EDA is to support the analysis of data prior to making any
conclusions. It may aid in the detection of apparent errors, as well as a deeper
understanding of data patterns, the detection of outliers or anomalous events, and
the discovery of interesting relationships between variables.

Content based filtering /popularity-based filtering

It is a type of Recommendation System which works on the principle of popularity
and or anything which is in trend. These systems check about the product or movie
which are in trend or are most popular among the users and directly recommend
those.
In the content based collaborative filtering we are finding the highest average rating
of each book. To find out the highest average rating we merge the ratings dataset
with books dataset
ratings_with_name=ratings.merge(books,on='ISBN')

Upon merging the rating dataset with book dataset we are grouping the ratings with
respect to the name of the book
num_rating_df=ratings_with_name.groupby('Book-Title').count()['Book
Rating'].reset_index()
num_rating_df.rename(columns={'Book-Rating':'num_ratings'},inplace=True)
num_rating_df

Using the mean function calculating the average ratings of the books.
avg_rating_df=ratings_with_name.groupby('Book-Title').mean()['Book
Rating'].reset_index()
avg_rating_df.rename(columns={'Book-Rating':'avg_rating'},inplace=True)
avg_rating_df

28
By sorting the average rating we can retrive the first 50 highest rating books fro the
dataset
popular_df=popular_df[popular_df['num_ratings']>=250].sort_values('avg_rating',a
scending=False).head(50)
popular_df

Collaborative based filtering Approach

Collaborative filtering is used by most recommendation systems to find similar
patterns or information of the users, this technique can filter out items that users like
on the basis of the ratings or reactions by similar users.

Collaborative filtering is a technique that can filter out items that a user might like on
the basis of reactions by similar users.

It works by searching a large group of people and finding a smaller set of users with
tastes similar to a particular user. It looks at the items they like and combines them
to create a ranked list of suggestions.

There are many ways to decide which users are similar and combine their choices
to create a list of recommendations. This article will show you how to do that with
Python.

To experiment with recommendation algorithms, you’ll need data that contains a set
of items and a set of users who have reacted to some of the items.

The reaction can be explicit (rating on a scale of 1 to 5, likes or dislikes)

or implicit (viewing an item, adding it to a wish list, the time spent on an article).

While working with such data, you’ll mostly see it in the form of a matrix consisting
of the reactions given by a set of users to some items from a set of items. Each row
would contain the ratings given by a user, and each column would contain the
ratings received by an item.

To build a system that can automatically recommend items to users based on the
preferences of other users, the first step is to find similar users or items. The second
step is to predict the ratings of the items that are not yet rated by a user.

29
x=ratings_with_name.groupby('User-ID').count()['Book-Rating']>200

padhe_likhe_users=x[x].index

filtered_rating=ratings_with_name[ratings_with_name['User-
ID'].isin(padhe_likhe_users)]

y=filtered_rating.groupby('Book-Title').count()['Book-Rating']>=50

famous_books=y[y].index

final_ratings=filtered_rating[filtered_rating['Book-Title'].isin(famous_books)]

pt=final_ratings.pivot_table(index='Book-Title',columns='User-ID',values='Book-
Rating')

Cosine Similarity

cosine similarity means the similarity between two vectors of inner product space, It
is measured by the cosine of the angle between two vectors.

from sklearn.metrics.pairwise import cosine_similarity

similarity_scores=cosine_similarity(pt)

similarity_scores.shape

def recommend(book_name):

#index fetch

index=np.where(pt.index==book_name)[0][0]

similar_items=sorted(list(enumerate(similarity_scores[index])),key=lambda
x:x[1],reverse=True)[1:6]

data=[]

for i in similar_items:

30
 item=[]

temp_df=books[books['Book-Title']==pt.index[i[0]]]

item.extend(temp_df.drop_duplicates("Book-Title")['Book-Title'].values)

item.extend(temp_df.drop_duplicates("Book-Title")['Book-Author'].values)

item.extend(temp_df.drop_duplicates("Book-Title")['Image-URL-M'].values)

data.append(item)

return data

Accuracy

One of the approaches to measure the accuracy of your result is the Root Mean
Square Error (RMSE), in which you predict ratings for a test dataset of user-item
pairs whose rating values are already known. The difference between the known
value and the predicted value would be the error. Square all the error values for the
test set, find the average (or mean), and then take the square root of that average
to get the RMSE.

Website Deployment

We are using the pycharm community to deploy the website. By creating the project
book recommendation System.

Flask provides configuration and conventions, with sensible defaults, to get started.
This section of the documentation explains the different parts of the Flask framework
and how they can be used, customized, and extended. Beyond Flask itself, look for
community-maintained extensions to add even more functionality.

def create_app():
app = Flask( name )
hello.init_app(app)
return app

31
 CHAPTER 6

RESULTS AND DISCUSSION

Today the World Wide Web provides users with a vast array of information, and
commercial activity on the Web has increased to the point where hundreds of new
companies are adding web pages daily. This has led to the problem of information
overload. Recommender systems have been developed to overcome this problem
by providing recommendations that help individual users identify content of interest
by using the opinions of a community of users and/or the user’s preferences.

The aim of this thesis was to design and evaluate different approaches for producing
personalised recommendations within the book domain. To achieve this goal, the
project first investigated existing recommender systems and profiling techniques.
The next step was to build users’ profiles by monitoring users’ behaviour, and
develop three different approaches for producing recommendations. Finally, an
evaluation of the system recommendations’ accuracy was done, by first conducting
live user experiments and then performing offline analysis to measure the
recommendations’ accuracy using appropriate methods for testing.

The system evaluation results show that the accuracy of the system
recommendations is very good and that a recommender system based on the
combination of content-based and collaborative filtering approaches provides more
accurate recommendations for the book domain.

32
 CHAPTER 7
CONCLUSION

7.1 CONCLUSION
All of our systems– purely content-based, purely collaborative-filtering, and hybrid–
performed quite well. Looking back on the project, one thing that we might have
chosen to do differently in retrospect would have been to spend more time searching
for a dataset of ratings with a higher rating variance per user. Had we been able to
find such a dataset, our implementations of algorithms would have been tested on
data that would have been more representative of what a typical commercial
recommendation system could access in creating its predictions. However, given
the data that was available to us, as well as the results our various approaches
produced, our systems were largely successful, providing insight into how the
different systems we regularly use work and the varying algorithms that make that
possible.

7.2 FUTURE WORK

Given more information regarding the books dataset, namely features like Genre,
Description etc., we could implement a content-filtering based recommendation
system and compare the results with the existing collaborative-filtering based
system.
We would like to explore various clustering approaches for clustering the users
based on Age, Location etc., and then implement voting algorithms to recommend
items to the user depending on the cluster into which it belongs.
7.3 RESEARCH ISSUES
One of the challenges faced by our research was the unavailability of reliable
training datasets. In fact, this challenge faces any researcher in the field. However,
although plenty of articles about predicting phishing websites using data mining
techniques have been disseminated these days, no reliable training dataset has
been published publically, maybe because there is no agreement in literature on the
definitive features that characterize phishing websites, hence it is difficult to shape
a dataset that covers all possible features.

33
In this article, we shed light on the important features that have proved to be sound
and effective in predicting phishing websites. In addition, we proposed some new
features, experimentally assign new rules to some well-known features and update
some other features.

7.3 IMPLEMENTATION ISSUES

• Handling of sparsity was a major challenge as well since the user interactions
were not present for the majority of the books.

• Understanding the metric for evaluation was a challenge as well.

• Since the data consisted of text data, data cleaning was a major challenge in
features like Location etc.

• Decision making on missing value imputations and outlier treatment was quite
challenging as well.

34
 REFERENCES
[1] Ahuja, Rishabh, Arun Solanki, and Anand Nayyar.” Movie recommender system
using K-Means clustering and K-Nearest Neighbor.” In 2019 9th International
Conference on Cloud Computing, Data Science Engineering (Confluence), pp. 263-
268. IEEE, 2019.

[2] Badriyah, Tessy, Erry Tri Wijayanto, Iwan Syarif, and Prima Kristalina. ”A hybrid
recommendation system for E-commerce based on product description and user
profile.” In 2017 Seventh International Conference on Innovative Computing
Technology (INTECH), pp. 95-100. IEEE, 2017.

[3] Chen, Junnan, Courtney Miller, and Gaby G. Dagher. ”Product recommendation
system for small online retailers using association rules mining.” In Proceedings of
the 2014 International Conference on Innovative Design and Manufacturing
(ICIDM), pp. 71-77. IEEE, 2014.

[4] Jisha, R. C., Ram Krishnan, and Varun Vikraman. ”Mobile applications
recommendation based on user ratings and permissions.” In 2018 International
Conference on Advances in Computing, Communications and Informatics (ICACCI),
pp. 1000-1005.IEEE, 2018.

[5] Keerthana, N. K., Shriram K. Vasudevan, and Nalini Sampath. ”An Effective
Approach to Cluster Customers with a Product Recommendation System.” Journal
of Computational and Theoretical Nanoscience Vol. 17, No. 1, pp. 347-352.IEEE,
2020.

[6] Kurmashov, Nursultan, Konstantin Latuta, and Abay Nussipbekov. ”Online book
recommendation System.” In 2015 Twelve International Conference on Electronics
Computer and Computation (ICECCO), pp. 1-4. IEEE, 2015.

[7] Kurup, Ayswarya R., and G. P. Sajeev. ”Task recommendation in rewardbased

crowdsourcing systems.” In 2017 International Conference on Advances in
Computing, Communications and Informatics (ICACCI), pp. 1511-1518. IEEE,
2017.

35
[8] Maya L Pai1,Suchithra M. S2andDhanya M, ”Analysis of Soil Parameters for
Proper Fertilizer Recommendation to Increase the Productivity of Paddy Field
Cultivation ”,Department of Computer Science and ITSchool of Arts and
Sciences,KochiAmrita Vishwa Vidyapeetham, International Journal of Advanced
Science and Technology Vol. 29, No. 03, pp. 4681-4696.IEEE, 2020

[9] M¨obius Data Scientist at Healthcare Melbourne, Victoria, Australia” Book

Recommendation Dataset Build state-of-the-art models for book recommendation
system”, kaggle, 2016.

[10] Mohamed, Marwa Hussien, Mohamed Helmy Khafagy, and Mohamed Hasan
Ibrahim. ” Recommender systems challenges and solutions survey.” In 2019
International Conference on Innovative Trends in Computer Engineering (ITCE), pp.
149-155. IEEE, 2019.

[11] Murali, M. Viswa, T. G. Vishnu, and Nancy Victor. ”A Collaborative Filtering

based Recommender System for Suggesting New Trends in Any Domain of
Research.” In 2019 5th International Conference on Advanced Computing
Communication Systems (ICACCS), pp. 550- 553. IEEE, 2019.

[12] PV Devika, K Jothisree, PV Rahul, S Arjun, Jayasree Narayan. “Book

Recommendation System”, 2021 12th International Conference on Computing
Communication and Networking Technologies (ICCCNT), 2021

[13] Tewari, Anand Shanker, and Kumari Priyanka. ”Book recommendation system
based on collaborative filtering and association rule mining for college students.” In
2014 International Conference on Contemporary Computing and Informatics (IC3I),
pp. 135-138. IEEE, 2014.

36
 APPENDIX

A. SOURCE CODE
Fig A.1 Refers to the implementation of the flask for the website deployment.
from flask import Flask,render_template,request
import pickle
import numpy as np

popular_df=pickle.load(open('popular.pkl','rb'))
pt=pickle.load(open('pt.pkl','rb'))
books=pickle.load(open('books.pkl','rb'))
similarity_scores=pickle.load(open('similarity_scores.pkl','rb'))
app=Flask( name )
@app.route('/')
def index():
return render_template('index.html',
book_name = list(popular_df['Book-
Title'].values),
author=list(popular_df['Book-
Author'].values),
image=list(popular_df['Image-URL-
M'].values),

votes=list(popular_df['num_ratings'].values),

rating=list(popular_df['avg_rating'].values)

)
@app.route('/recommendation')
def recommendation_ui():
return render_template('recommendation.html')
@app.route('/recommend_books',methods=['post'])
def recommend():
user_input=request.form.get('user_input')
index = np.where(pt.index == user_input)[0][0]
similar_items = sorted(list(enumerate(similarity_scores[0])),
key=lambda x: x[1], reverse=True)[1:6]
data = []
for i in similar_items:
item = []
temp_df = books[books['Book-Title'] == pt.index[i[0]]]
item.extend(temp_df.drop_duplicates("Book-Title")['Book-
Title'].values)
item.extend(temp_df.drop_duplicates("Book-Title")['Book-
Author'].values)
item.extend(temp_df.drop_duplicates("Book-Title")['Image-
URL-M'].values)
data.append(item)
print(data)
return render_template('recommendation.html',data=data)

if name ==' main ':

app.run(debug=True)

Fig A.1: Flask Sourcecode

37
 B. SCREENSHOTS

First of all we are importing the required libraries and datasets (Fig B.1)

Fig B.1 Importing Libraries and Datasets

38
In the next step data pre-processing is carried out to modify the data as required
(FigB.2)

Fig B.2 Data Pre-processing

39
In the Fig B.3, the books dataset is merged with ratings dataset to evaluate the
highest average rating of the books

Fig B.3 Popularity based filtering/ Content based filtering Model

Fig B.4 describes the output of the content based filtering model i.e. the top 10
highest average rating books are displayed

Fig B.4 Popularity based filtering / Content based filtering output

40
Collaborative filtering model recommends the books to users based on the books
interacted by the users(fig B.5)

Fig B.5 Collaborative Based Filtering model Output

RMSE and MAE values are shown in Fig B.6

Fig B.6 RMSE and MAE valuesw

41
Webpage deployment of content based filtering model is shown in Fig B.7

Fig B.7 Webpage for content-based filtering Model

42
Webpage deployment of collaborative based filtering is shown in the fig B.8

Fig B.8 Webpage for collaborative based filtering model

43
 C. RESEARCH PAPER

Book Suggester System using Machine Learning

Bavisetti Gowtham Ajay Betha Shanmukh Satyadev Dr. M.D. Anto Praveena
Computer Science and Engineering Computer Science and Engineering Computer Science and Engineering
Sathyabama Institute of Science and Sathyabama Institute of Science and Sathyabama Institute of Science and
Technology Technology Technology
Chennai, India Chennai, India Chennai, India
bavisettiajay9@gmail.com shanmukhsatyadev@gmail.com antopraveena.cse@sathyabama.ac.in

Abstract—As human civilization and web technology I. INTRODUCTION

become advanced, more and more data is being
gathered by online users. This ocean of information Recommender Systems (RSs) are essential for
provides a wide variety of resources. The question of increasing the e-commerce system's revenue. In recent
how to investigate the utilization of these resources has years, RS has been used in many different fields. Its first
become urgent. These data sets can be analyzed to concentration was on the entertainment and online retail
extract hidden information or finding new industries. Its use case is being developed in a variety of
information. Data mining is a technique for obtaining fields, including learning online, net banking, booking
usable information from computers and technological tickets online, medication, social media, and others.
methods. A typical tool in today's information age is
the search engine. Every day, people utilize search Customers find it difficult to purchase books from a
engines like Google, Baidu, Bing, and others. Yet, due large selection of books on an e-commerce website. The
to its universality, they are unable to provide RS is the best option to reduce the overhead issue.
consumers with accurate search results that are because it enables the user to find an ideal book based
tailored to their special needs and histories. In on their interests. Our primary goal for research is to
response, it suggests the idea of customized built and develop a suggestion of books for reading books in
customized suggestion systems of creating a dual online.
relationship among the user and products of
information, utilizing current process of choosing or a The twenty-first century is here. Nearly everything in
familiar connection, quarrying similar suggestions of the world uses internet service these days. IoT is also
interest of every user, and carry out customized identifying growth, which means that the most
suggestions. The collaborative recommendation important objects in this world are digitalized.
algorithm will be studied in this article along with a k- eLearning is crucial, particularly in the COVID 19
means clustering and k-nearest neighbor method that pandemic. The discipline of eLearning benefits greatly
is suggested. from the use of recommendation systems. The most
beneficial is that eLearning offered to end customers
Keywords—Data mining, collaborative filtering, was "Response based on the demand" that signifies the
machine learning, and k-means clustering, k-nearest appeal of the client will be fulfilled. Because learning
neighbor. online includes variety of portable devices (including
laptops, tablets, and smartphones).

44
 suggestion of books for online learning is illustrated in
The system provides prompt recommendations for any Section IV and the words are briefly defined in Section
services and goods the customer requests "Whatever V. The dataset, experimental evaluation, and result
and Whenever" thanks to these portable technologies. analysis are all described in Part VI. The section of the
Having a observation on the scenario when a user essay that concludes is Section VII.
wishes to choose an e-book or other reading content but
lacks the necessary personal experience. Unfortunately,
because there were so many things or e-resources II. LITERATURE REVIEW
available, they were unable to find the right ones. The
system that suggests is what we have suggested to avoid The field of recommendation system has expanded
this situation. Fig.1 depicts the flow of the recommender significantly over the past 20 years. Customers have
system overloaded by the information which is addressed by
offering unique, specialized material and repair advice.
A few modern cutting-edge techniques are shown in the
analysis sections.

Proposal frameworks deal with providing relevant

advice in surprisingly intelligent ways. Typically,
association rules, content-based filtering, and
collaborative filtering are used. Moreover, a
straightforward recommendation system for mobile
applications is created without sacrificing rating, size,
or permission features. There are two models provided,
one based on worker incentive features and the other on
worker job features. Using real-world datasets, the
Fig. 1. Flow of a Recommender System suggest strategy is compared to several connected
methods.
RS was first introduced by Resnick and Varion [1] for
developing a collaborative filtering method [2]. Users decide on a set of data through collaborative
Eventually, several researchers utilized the phrase and filtering. When creating the user want facts set to
gave it different definitions. On the internet, numerous represent the different facets of the user's profile, special
recommendation systems are presented and used in characteristics are taken into consideration. It is a
different sectors, and each one aims to accurately technique for providing users with up-to-date hints from
anticipate in accordance with user interest. Several a dataset based on how closely their side interest profiles
researchers also looked into the novelty, privacy, resemble another user. It is quite tough to side-features
scalability, accuracy, usability, and other aspects of RS. for a product or query, and clean that technology cannot
Many methods, including collaborative filtering, be controlled. Customers' information is used by a
knowledge-based, demographic, and hybrid methods, content-based recommender, either openly or through
can be used to make recommendations. verification.

Collaborative Filtering (CF) is the method for creating

recommendations that is most generally used and A user profile is created using this data, and it is then
approved [3]. The CF technique finds comparable utilized to provide recommendations to the user. The
products that other users with similar interests favor algorithm recognizes customer’s various interests and
based on prior interactions (rating, remark) [4]. Based make suggestions based on the interests of the similar
on profile attributes, content-based filtering determines user interests. The capacity of this paradigm to increase
the user's choice based on the items' contents [5]. the users' present benefits is confined. A huge data set
Knowledge-based works rely on knowledge expertise of things can be explored using association rule mining
regarding the user and the item and suggest product to uncover fascinating associations and interactions
based on which items are connected to the user's between the products. It is a technique for searching for
interests. When suggesting an item, Demographic RS recurrent patterns, parallels, or similarities between
bases its operation on user demographic information datasets from different databases, including relational
from their profiles (such as gender, age, location, etc.). databases, transactional databases, and various kinds of
repositories.
This article sections are structed as follows : A literature
Review on recommender systems is included in Part II. The foremost weakness of the prior methods is the only
many similarity methods engaged in the RS are address one application, namely, an existing client
presented in Section III. The suggested model of the receiving recommendations from the dataset, and do not

45
address usage scenarios like proposing a new book to a dataset contains the user ID, location, age. The third
new customer or a generic counsel based on the dataset. dataset is the ratings dataset. The ratings dataset
Our system includes each of these programs, giving contains the user ID, ISBN of the book and the ratings
users a selection of options for picking the ideal book to provided by the users for the particular book.
read.
B. Algorithm
The primary shortcoming of the earlier solutions is that
they only handle one application, namely, a current 1) Pearson correlation: To gauge the linear correlation
customer receiving suggestions from the dataset and do between variables, co-efficient of Pearson R correlation
not deal with use scenarios such as recommending a is utilized. Its purpose is to build a recommendation
new book to a brand-new customer, or a generic advice using rating counts.
based on the dataset. Each of these applications is
included in our system, providing customers with a depending upon the ratings and the count of rating, a
variety of options for selecting the appropriate book to graph of rating distribution is drawn, and most
read. individuals have given ratings of 0.

III. METHODOLOGY In order to obtain the count in desc order, first group the
rating data frame by ISBN, then take the book rating
Algorithm used in "Book Recommendation column. Hence, if more people rate a book, it must be a
System" initiative aims to assist users in selecting the very well-known book overall. In order to do that, we
right choice of books that piques their interest and so have created a data frame using all the ISBN numbers,
motivate them to learn more. In this case, we're using then combined it with the books dataset's most popular
Cosine similarity, KNN, and Pearson correlation. titles using the ISBN field. Therefore, in conclusion, we
Seeing similarities across the books is becoming more can find the top 5 books based on the number of ratings
and more routine. As previously indicated, the major here. The ratings mean as well as the ratings count were
three use cases for this system are suggestions for discovered for the correlation. Then, present the data in
current users, recommendations for new users, and desc order after generating two fields in data frames:
ratings for newly uploaded books. Several approaches rating of the book and rating count. As a result, if we are
are used to deal with each of these. The main method using the rating count to determine popularity and my
used in this project is collaborative filtering based on book does not have a decent rating but the greatest
users. Depending upon the ratings given to the product number of people have given it a rating, it cannot be a
by another reader who share the target user's widely read book, so we should not promote it. In order
preferences, the system predicts what a user will like. for them to make recommendations, they must take into
Fig 2 depicts the difference between the content based account the rating average and the amount of ratings.
filtering and collaborative based filtering. We have some factual importance where users with less
than 200 evaluations and books with fewer than 100
evaluations are disallowed in contemplation to create
the best recommendation system. At that time, having
the opinion to combine user id and ISBN using the pivot
table; therefore, by using the indexes as user id on the
column features, it would show if the individual has
provided any ratings or not. When the pivot is used, the
rating table is actually transformed into a 2D matrix if
the user has not provided any ratings, which is basically
represented as nan. Then a correlation between the
ratings and the ratings average is discovered. Fig 3
depicts the output of pearson algorithm.

Fig. 2. Content based vs Collaborative based

Filtering
A. Dataset Description
For the dataset, three csv files are taken. First one is the
books file. The attributes of the books dataset are ISBN,
Book title, Book author, year of publication, publisher,
image URL small, image URL medium, image URL
large. The second dataset is user’s dataset. The user’s

46
 unsupervised learning algorithm is nearest neighbor. It
is imported as sklearn.neighbors from the sklearn
library. When determining users who are similar, the
user's previous ratings of the books are considered.

Three new features were developed. One to find the K-

Nearest Neighbors, second is to forecast the average
ratings of a specific book, and the third is to recommend
the user the best-rated books. This information is
transformed into a cross tabulation with ISBN columns
and the id of the user as the index, and all 0 ratings are
eliminated. The data is essentially represented as a
matrix in the pivot table. It is determined how many
Fig. 3. Output of the Pearson algorithm books each user has reviewed. The system receives a
user's distinctive ID, a distinctive ISBN of book, the
2) KNN: KNN is a machine learning estimator used to number of products the reader has reviewed, and the
find set of alike readers depending upon basic book cross tabulation of ratings.
ratings and provide data using the average rating of the
top k nearest neighbors. KNN is a machine learning The standard deviation and the user ratings mean are
estimator used to find groups of similar users based on used to calculate the book ratings. The average of all the
basic book ratings and provide data utilizing the average user-rated books is represented by the mean rating. The
rating of the highest k nearest neighbors. weighted average is calculated using both the mean rate
of the mean of the comparable readers of the given book
In this suggestion system, similar readers are found and the measure of their similarity. The distance
using similarity of cosine by converting the table to a produced via K Nearest Neighbor is used to determine
two-dimensional matrix and then filling in the gaps with the similarity value.
NULL. The network data frame’s values (ratings) were
then converted into a SciPy sparse grid for more Prediction Rate = Rate of the Mean + Standard
accurate computations. Finally, this computation will Deviation.
determine the similarity of the cosine between rating
vectors. Unsupervised techniques with The user is then shown the top 10 books after the ratings
sklearn.neighbors are used to get the Closest Neighbors. are organized in descending order. If those projected
ratings are fewer than 6, a message asking the user to
To determine how comparable two items are, one can examine some general suggestions determined by
use the cosine similarity measurement. In terms of Pearson correlation is displayed. Fig 4 depicts the output
numbers, it calculates the intersection cosine of two of collaborative Filtering output
vectors that are visualized into a multidimensional
space. The yield value ranges from 0 to 1. (0 denotes the
absence of comparability, while 1 suggests that both
elements are equally rated at 100.

A collaborative filtering recommendation system based

on items is developed using KNN. The five books with
the highest similarity scores to the selected book are
listed. Since similarity is determined by distance, the
higher the value, the less similarity there is. Following
that, they are put in ascending order. When a user types
in their favorite book from the dataset, the algorithm
suggests the user's next five most related novels. Fig. 4. Collaborative Filtering output

3) Collaborative filtering: Collaborative filtering is one 4) Suggestion of books to the New Users: A few books
of the methods we use to forecast the book ratings from should also be recommended to any new users who need
an already subscribed user by computing the to be added to the data set so that they can rate them.
resemblance among readers. The fundamental concept Future recommendations for that person can be
of collaborative filtering is this. Let us say we want to enhanced using this rating. The top 10 books, as
make recommendations for user x. To start, we identify determined by the total average rating of all the books,
a lot of other users that have user x's likes and dislikes. are presented to each new user who has been added to
K Nearest Neighbor is used to accomplish this. An this recommendation system. By dividing the total

47
number of ratings for each book by the total number of approach for determining a model's error.
ratings, the average ratings are determined. Afterwards, 2) MAE: The only difference between this method and
they are arranged in the order of decrement. Following the former is that with MAE, In the case of linear
the users' ratings of these books, collaborative filtering continuous variables, the error is estimated.
is used to create the users' subsequent suggestions. Fig
5 depicts the books with highest rating and fig 6 depicts From the ratings pivot table five readers were chosen
the highest rated author’s randomly. The five users anticipated and saved ratings
of the previously reviewed books. Also, each of the
dataset's actual evaluations was saved. There were 666
total ratings. These served as a basis for calculating the
Root Mean Squared Error and Mean Absolute Error
values.

From the Fig.7 Root Mean Squared Error value is equal

to 1.836 and Mean Absolute Error values is equal to
0.976.

Fig. 7. RMSE and MAE

Fig. 5. Highest average rated books

V. CONCLUSIONS AND FUTURE WORK

This research suggests a machine learning algorithm

named collaborative filtering mechanism for a
recommendation of books. By giving book
recommendations, people's reading habits improve,
which boosts their vocabulary, expertise, and
knowledge. The data gathered from reader reviews of
completely unrelated novels is utilized. The dataset has
an excessive several books and a wide variety of
consumers. Our designed system makes the most of the
information's unique alternatives to provide a speedy
response and high-caliber recommendations.
Fig. 6. Highest-rated authors in terms of average
Collaborative Filtering is the algorithm used by the
system to generate recommendations. The cosine
5) Inserting New Book: To be considered for a user
similarity approach is used to precisely quantify the
suggestion and to be part of the data set, a new book
similarities between the users. Based on the average
needs to have a rating. The average ratings of all the
ratings calculated and gathered from the various users
author's other books that are already part of the data set
the top rated gooks are suggested for the book readers.
are used to compute this score. According to their
author, books are sorted. After adding up all of the
Time constraints are the biggest drawbacks of this
ratings for each book, a percentage of the total ratings is
machine learning model named collaborative filtering.
calculated. It is also possible to see how many people
Collaborative filtering consumes more time to suggest
read the book by that author by looking at the overall
books to the readersbecause of the vastness of the
rating count. The new book that was added to the data
dataset. Usage of the supervised instruction, this can be
collection is then given this rating as its initial rating.
easily resolved. If the customer seeking for a book to
read, our suggestion system may be just what you are
IV. EXPERIMENTAL SETUP AND RESULTS
looking for. It offers complete satisfaction.
A. Performance Assessment
REFERENCES
Two measures are used to evaluate the collaborative
filtering's performance: -
[1] Kurmashov, Nursultan, Konstantin Latuta, and
1) RMSE: RMSE is a widely used and accepted
Abay Nussipbekov. ”Online book recommendation

48
System.” In 2015 Twelve International Conference on Recommendation to Increase the Productivity of Paddy
Electronics Computer and Computation (ICECCO), pp. Field Cultivation ”,Department of Computer Science
1-4. IEEE, 2015. and ITSchool of Arts and Sciences,KochiAmrita
Vishwa Vidyapeetham, International Journal of
[2] Tewari, Anand Shanker, and Kumari Priyanka. Advanced Science and Technology Vol. 29, No. 03, pp.
”Book recommendation system based on collaborative 4681-4696.IEEE, 2020
filtering and association rule mining for college
students.” In 2014 International Conference on [11] Jisha, R. C., Ram Krishnan, and Varun Vikraman.
Contemporary Computing and Informatics (IC3I), pp. ”Mobile applications recommendation based on user
135-138. IEEE, 2014. ratings and permissions.” In 2018 International
Conference on Advances in Computing,
[3] Mohamed, Marwa Hussien, Mohamed Helmy Communications and Informatics (ICACCI), pp. 1000-
Khafagy, and Mohamed Hasan Ibrahim. ” 1005.IEEE, 2018.
Recommender systems challenges and solutions
survey.” In 2019 International Conference on
Innovative Trends in Computer Engineering (ITCE), [12] Keerthana, N. K., Shriram K. Vasudevan, and
pp. 149-155. IEEE, 2019. Nalini Sampath. ”An Effective Approach to Cluster
Customers with a Product Recommendation System.”
[4] Badriyah, Tessy, Erry Tri Wijayanto, Iwan Syarif, Journal of Computational and Theoretical Nanoscience
and Prima Kristalina. ”A hybrid recommendation Vol. 17, No. 1, pp. 347-352.IEEE, 2020.
system for E-commerce based on product description
and user profile.” In 2017 Seventh International [13] PV Devika, K Jothisree, PV Rahul, S Arjun,
Conference on Innovative Computing Technology Jayasree Narayan. “Book Recommendation System”,
(INTECH), pp. 95-100. IEEE, 2017. 2021 12th International Conference on Computing
Communication and Networking Technologies
[5] Ahuja, Rishabh, Arun Solanki, and Anand Nayyar.” (ICCCNT), 2021
Movie recommender system using K-Means clustering
and K-Nearest Neighbor.” In 2019 9th International
Conference on Cloud Computing, Data Science
Engineering (Confluence), pp. 263-268. IEEE, 2019.

[6] Murali, M. Viswa, T. G. Vishnu, and Nancy Victor.

”A Collaborative Filtering based Recommender System
for Suggesting New Trends in Any Domain of
Research.” In 2019 5th International Conference on
Advanced Computing Communication Systems
(ICACCS), pp. 550- 553. IEEE, 2019.

[7] Chen, Junnan, Courtney Miller, and Gaby G.

Dagher. ”Product recommendation system for small
online retailers using association rules mining.” In
Proceedings of the 2014 International Conference on
Innovative Design and Manufacturing (ICIDM), pp. 71-
77. IEEE, 2014.

[8] M¨obius Data Scientist at Healthcare Melbourne,

Victoria, Australia” Book Recommendation Dataset
Build state-of-the-art models for book recommendation
system”, kaggle, 2016.

[9] Kurup, Ayswarya R., and G. P. Sajeev. ”Task

recommendation in rewardbased crowdsourcing
systems.” In 2017 International Conference on
Advances in Computing, Communications and
Informatics (ICACCI), pp. 1511-1518. IEEE, 2017.

[10] Maya L Pai1,Suchithra M. S2andDhanya M,

”Analysis of Soil Parameters for Proper Fertilizer

49
50