A CASE STUDY REPORT ON

ON
NETFLIX RECOMMENDATION SYSTEM

BACHELOR OF ENGINEERING (INFORMATION

TECHNOLOGY) SUBMITTED BY

Miss.S w a t i Z a g a d e Exam No:

Miss.Nikita Borse Exam No:
Miss.Vibhuti Gudahghe Exam No:
Miss.Akshada Gahile Exam No:

Under The Guidance of

Dr.V.R.Sonawane

Department of Information Technology

MVPS K.B.T College of Engineering, Nashik

2022-23
 Abstract

In pursuit of this objective of mobile recommendation, the paradigms of

machine learning and natural language processing are one of the best paradigms.
The products can be properly examined and realized for their prospective for
suggestion to a specific user by implementing these ways. The natural language
processing method is particularly essential since it can examine customer reviews
in order to locate helpful products with favorable ratings which provide the cus-
tomer with appropriate recommendations. Machine learning algorithms can assist in
determining sarcasm, which can be extremely useful in automatically iden- tifying
favorable and unfavorable evaluations. This approach utilizes the user requirements
and a dataset along with NLP in the form of Bag of Words and TF- IDF, in addition
to Fuzzy Artificial neural Networks and Collaborative filtering to achieve precise
Mobile recommendations. The recommendation system has been subjected to a round
of experimentation and the outcomes have been displayed with favorable results.

Keywords: Natural Language Processing, Bag of Words, TF-IDF, Fuzzy ANN,

Collaborative Filtering.
 Contents

ACKNOWLEDGMENT I
ABSTRACT II
LIST OF FIGURES VI
LIST OF TABLES VII

Chapter Chapter Name Page

No. No.

1 INTRODUCTION 1
1.1 Background ..................................................................................................... 1
1.2 Relevance ........................................................................................................ 3
1.3 Organization of the Project Report........................................................... 3

2 LITERATURE SURVEY 5

3 PROJECT STATEMENT 9
3.1 Problem Definition .................................................................................... 9
3.2 Proposed Methodology ................................................................................... 9

4 SOFTWARE REQUIREMENTS AND SPECIFICATION 10

4.1 Introduction ................................................................................................... 10
4.1.1 Project Scope .................................................................................... 10
4.1.2 User Classes and Characteristics ................................................ 11
4.1.3 Assumption and Dependencies ........................................................ 12
4.2 FUNCTIONAL REQUIREMENTS ......................................... 12
4.2.1 Preprocessing .................................................................................... 12
4.2.2 Bag of Words ................................................................................... 13
 4.2.3 TF-IDF . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
.
4.2.4 Fuzzy ANN . . . . . . . . . . . . . . . . . . . . . . . . . 13
. .
4.2.5 Collaborative Filtering . . . . . . . . . . . . . . . . . . . . 14
.
4.3 EXTERNAL INTERFACE REQUIREMENTS . . . . . . . 14
. . . . .
4.3.1 User Interfaces . . . . . . . . . . . . . . . . . . . . . . . . 14
.
4.3.2 Hardware Interfaces . . . . . . . . . . . . . . . . . . . . . . 15
.
4.3.3 Software Interfaces . . . . . . . . . . . . . . . . . . . . . . 15
.
4.3.4 Communication Interfaces . . . . . . . . . . . . . . . . . . 15
.
4.4 Nonfunctional Requirements . . . . . . . . . . . . . . . . . . . . . 15
.
4.4.1 Performance Requirements . . . . . . . . . . . . . . . . . . . 15
4.4.2 Safety Requirements . . . . . . . . . . . . . . . . . . . . . 16
.
4.4.3 Security Requirements . . . . . . . . . . . . . . . . . . . . 16
.
4.4.4 Software Quality Attributes . . . . . . . . . . . . . . . . . 16
.
4.5 System Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 16
.
4.5.1 Software Requirements(Platform Choice) . . . . . . . . . . . 16
4.5.2 Hardware Requirements . . . . . . . . . . . . . . . . . . . 16
.
4.6 Analysis Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
.
4.7 System Implementation Plan . . . . . . . . . . . . . . . . . . . . . 18
.

5 SYSTEM DESIGN 19
5.1 System Architecture ................................................................................ 19
5.2 System Design ......................................................................................... 19
5.3 Data Flow Diagrams ............................................................................... 20
5.3.1 DFD Level 0 ................................................................................ 20
5.3.2 DFD Level 1 ................................................................................ 20
5.3.3 DFD Level 2 ................................................................................ 21
 5.4 UML Diagrams ........................................................................................ 21
5.4.1 Use Case Diagram ....................................................................... 21
5.4.2 Activity Diagram ......................................................................... 22
5.4.3 Sequence Diagram............................................................................ 22
5.4.4 Component Diagram ................................................................... 22
5.4.5 Deployment Diagram .................................................................. 23
5.4.6 Package Diagram ......................................................................... 24
5.4.7 State Transition Diagram ............................................................ 24

6 IMPLEMENTATION 25
6.1 Implementation Detail ............................................................................. 25
6.2 Algorithm Details .................................................................................... 29
6.3 Tools and Technologies Used ...................................................................... 29

7 SOFTWARE TESTING 31
7.1 Type of Testing Used ................................................................................... 31
7.1.1 Performance Testing ........................................................................ 31
7.1.2 System Testing ............................................................................ 31
7.1.3 Recovery Testing ......................................................................... 31
7.1.4 Security Testing ........................................................................... 31
7.1.5 Integration Testing....................................................................... 32
7.2 Test Cases ................................................................................................ 32

8 RESULTS 33
8.1 Snapshot ........................................................................................................ 33
8.2 Results and Discussion............................................................................ 36

9 CONCLUSION 39
9.1 Conclusion .................................................................................................... 39
9.2 Future Scope ................................................................................................. 39

Appendix A REFERENCES 41
 List of Figures

3.1 System Diagram ........................................................................................ 9

4.1 System Plan ............................................................................................. 17

4.2 Timeline Chart ........................................................................................ 18
5.1 System Architecture ................................................................................ 19
5.2 System Design ......................................................................................... 19
5.3 DFD Level 0 Diagram ...................................................................... 20
5.4 DFD Level 1 Diagram ...................................................................... 20
5.5 DFD Level 2 Diagram ...................................................................... 21
5.6 Use Case Diagram ................................................................................... 21
5.7 Activity Diagram.................................................................................... 22
5.8 Sequence Diagram ....................................................................................... 22
5.9 Component Diagram ............................................................................... 23
5.10 Deployment Diagram .............................................................................. 23
5.11 Package Diagram ..................................................................................... 24
5.12 State Transition Diagram ........................................................................ 24

6.1 System Overview ......................................................................................... 25

8.1 Login Frame ...................................................................................... 33

8.2 Registration Frame .................................................................................. 33
8.3 User Operation Frame ............................................................................. 34
8.4 Edit Profile Frame .................................................................................. 34
8.5 Search Frame ................................................................................................ 35
8.6 Output Frame ........................................................................................... 35
8.7 View History Frame................................................................................ 35
8.8 : Precision and Recall Measurement for the Mobile recommendation 37
8.9 Comparison of Precision and Recall for the Mobile recommendation 37
4.1 Hardware Requirements ............................................................................... 17

7.1 Test Cases for User ................................................................................. 32

 CHAPTER 1

INTRODUCTION

1.1 Background

Technological advancement has been very conducive to a large section of society

as it has been useful in providing the necessary growth in various different sectors. This
has been useful in achieving immense strides in the growth of the human civilization as
a whole. There has been an increased interest in these endeavors as more and more
individuals have been contributing to the advancements. This has been one of the
driving factors for success of the humans on this planet which has helped us survive
for such a long period of time. The increase in the survivability has also led to
improving lifestyle of human beings across the world.
The technology has been useful in the development of the internet platform.
The internet platform has been useful in determination of the various different ser- vices
and approaches. The initiation of the internet platform and its proliferation has
allowed for an increased utilization of the applications and other webpages on this
platform. The most innovative and highly useful approach is the inclusion of Electronic
commerce websites. These websites have been increasing consistently in the past few
years which has been due to the increased convenience offered by these platforms.
In recent years, the E-Commerce sector has experienced tremendous expan- sion
in online platforms. And that can be due to advancements in the online medium, which
have resulted in a huge rise in the customer base, making ma- jority Web service
implementations extremely cost-effective. This is vital for the implementation of the
information infrastructure, which has focused on improving the usefulness and
accessibility of the product’s consumers. The huge reduction in the cost of internet-
enabled gadgets has accelerated the expansion, resulting in a massive expansion in
the E-Commerce industry.

1
 There has been a large number of studies into how to strengthen this service
much further and supply consumers with innovative and exciting factors that
characterize it apart from its contemporaries. As a result, there has been an
abundance of research towards enhancing the customer experience and providing
distinctive properties that can engage a bigger percentage of subscribers. This enables
the service to expand and then provide new services, making it far more accessible
for consumers to purchase from the convenience of their own residences.Recommender
systems predicated on the user’s requirements and past pur- chasing history are an
example of a beneficial deployment. That would be a very valuable and successful
method for adding an additional dimension to the service quality provided to each
consumer on the E-Commerce website. Recommender systems are incredibly
sophisticated and necessitate a comprehensive understand- ing of the customer’s
buying habits as well as the numerous items that have been originally purchased that
are unique to that individual. This results in a large number of possibilities, which
can be intimidating and make calculating sugges-
tions incredibly challenging.
As a result of the deployment of the bag of words technique in this method-
ology, natural language processing was already established. The bag of words is
little more than a customized lexicon associated with the implementation. The
different bags of words are employed in our method to separate the favorable and
unfavorable terms in the assessments. This enables us to evaluate whether the
feedback is fully favorable or negative in order to assess the performance of the
commodity. This technique is further strengthened by the implementation of phrase
frequency and inverse document frequency, which can be used to identify the
relevance of a specific word and it can be beneficial in removing sarcasm from
customer reviews.
The use of a Fuzzy artificial neural network enables for the classification of
customer reviews into five components of positivity and negativity. This provides a
commodity to be labelled with a variety of labels, which could also aid in the rec-
ommendation process based on the required specification. This is also augmented by
the implementation of collaborative filtering, which efficiently and successfully

2
serves to notify the cosine similarity among the couple elements, the initial of thatis
previously suggested products to users with similar, and the second of these is the
suggestion supplied by our system. These two aspects are successfully consoli- dated
to substantially enhance the accuracy of the product recommendation. This research has
successfully expanded the technique, as well as an adequate assess- ment of the
evaluation metrics, which has been successful and efficiently beneficial, as well as
future research directions.

1.2 Relevance

E-commerce platforms have been getting increasingly popular with a large number
of people utilizing this platform increasingly. This is due to the increased convenience
and ease of use offered by these platforms in shopping for essential items and other
products at the comfort of your home. This is allowed a lot of disabled individuals
and individuals with mobility issues to effectively stay athome and order various items
that are needed easily. This has also led to a large number of e-commerce platforms that
are vying for the attention of the users and trying to increase the user base to be able to
provide effective solutions with ease. One such implementation is the realization of
product recommendations to theuser based on the various product reviews that are
posted on the products. This is a highly complex task as a large number of reviews are
posted for a product with varying degrees of positivity or negativity. Therefore to
achieve effective product recommendation an effective approach has been outlined in
this research article through the utilization of machine learning approaches. The
presented technique implements the Bag of Words approach, Term Frequency, and
Inverse Document Frequency along with Fuzzy Artificial Neural Network and
Collaborative Filtering to achieve highly accurate product recommendations. The
approach has been qualified through extensive experimentation of precision and recall
metrics which indicate very high accuracy of execution of this recommender system.

1.3 Organization of the Project Report

Following sections are dedicated as the further steps of the report

3
Section 2: Literature Survey
Section 3: Problem Statement
Section 4: System Requirement & Specification
Section 5: System Design
Section 6: Implementation
Section 7: Software Testing
Section 8: Results
Section 9: Conclusion
Section 10: References

4
 CHAPTER 2

LITERATURE SURVEY

This section of the literature survey eventually reveals some facts based on
thoughtful analysis of many authors work as follows.
Eriko Shibamoto [1] explains that taking images has become better in the recent
years due to the advancements and the improvements in the image capturing
technologies. There have been incremental updates that have been focusing on the
realization of handheld and smartphone cameras. Nowadays there are numerous
techniques that have allowed the inclusion of a camera in a smartphone and almost all
smartphones have a camera now. Therefore, the authors in this approach have been
focusing on the aspect of clicking images on the smartphone and allowing the
realization of an effective recommendation system based on the behavior of the
users.
Fenfang Xie [2] states that the smartphone have been proliferating in the recent
years. This can be owing to the increased conveniences offered by the smartphones for
each of the individuals. This has been evident in the increased use and the
manufacturing of smartphones that have been highly influential in the realization of
a large number of use cases for the smartphones. The increased adoption of the
smartphones have been crucial in the development of the various applications for
enabling increased functionality for the user. There are numerous applications on the
app store but the process of application recommendation can be improved further. For
this purpose the authors in this publication have proposed the useof graph based
neural networks to recommend apps to the user based on their interests.
Rui Liu [3] expresses that the increasing popularity of the smartphones have
been one of the defining features of this decade and have become almost universal in
its application and usage. The increased usage of smartphones has been critical towards
the development of better hardware and other features such as applica-

5
tions on the smartphone platform. Due to the constant usage of the devices there is a
lot of personal and private information on these devices that needs to be ef- fectively
secured. To achieve effective privacy and security of the user’s data on the
smartphones the authors have proposed the use of a crowdsourcing approach based on
recommendations. This approach identifies the user preferences which are used to
secure the data stored on the device and manage the permissions.
Dimas Indiarto Sumiko [4] narrates that the social media has turned into a necessity
nowadays with a large amount of reliance that is being put on these web services for
the purpose of communication. These web services such as social media networks have
been easy to use and convenient to achieve the communication with your loved ones
and family members. The social media outlets have a large base of regular users that
have been utilizing their services every single day. These social media networks have
been also providing the effective advertisements and other marketing strategies to the
users based on their preferences and other interests. This study details the customer
behavior with impulsive smartphone buying and correlating it with the YouTube
product recommendations.
Jinyang Liu [5] individuals all around the globe are already embracing smart-
phones currently, and a large number of consumers are employing smartphones
with Android operating systems and services. A numerous individuals are en-
gaging on the Android platform to develop and construct innovative solutions, but
their effort is fraught with challenges: completing a script document, or per- haps
even simply a code succession, is a challenging task which requires extensive
engineering expertise and interpretation. As a result, designing and developing
computer-based frameworks that assist programmers in their job has become a
major study area. Because API utilization is such an important component of
Android development, this article delves into the challenge of proposing API uti-
lizations to developers.
Pulkit Rustgi [6] highlights the existing authorization mechanism for Android
OS confidentiality protection, which is primarily reliant on individuals. Before or
after a software is implemented, users must determine what privileges it has per-
mission to. Application are welcome to answer users because for some materials

6
they believe they require. Authorization sought by these programs are frequently
provided to users via a dialogue box, with very few data supplied. This has been
ineffectual because consumers might not even fully comprehend the commodity to
which an application is asking access and/or simply lack the resources to investi-
gate the consequences of accepting a clearance query.
Yudai Kato [7] depicts the situation notwithstanding the significant advances in
information technology, anybody may readily communicate and exchange knowl-
edge irrespective of distance or location, and access a wealth of diverse material
via the World Wide Web. The equivalent may be true for visitor attractions.
Travelers must choose vital information from the large volume of tourist material,
and finding knowledge matched to their needs is quite challenging. As a result, by
using World Wide Web to supply travelers with useful sightseeing data is es-
sential. The authors have recommended using an effective technique that employs the
regularity of tourist visits for the objective of proposing a tourism sightseeing place
for this objective.
Nesrine Kadri [8] have created a healthier suggestion model based on smart-
phone sensor measurements. Drawing on WHO standards, this approach assesses the
condition of moderate physical activity for mobile phone users and determines whether
they are productive or not. The data was categorized into five groups by the
researchers, which included vigorous physical activities that used calories andother
sedentary activities. Machine learning methods outperformed deep learn- ing
systems in categorization. The challenge of uneven groupings throughout the
training process might be used to justify this. Researchers can use the suggestion
system once a week as a points of view. Researchers may further enhance the
findings and using another sensor smartphone information, the Global Navigation
Satellite system, to determine the kilometers range of vigorous exercise such as
walking and running.
Sushmita Roy [9] offers a movies recommendation engine consisting of multi-
ple Python language techniques. It accepts a collection of data containing various
movies published in recent years as source and generates suggestions based on sev- eral
methodologies. All of the recommender systems that have been adopted all

7
had their own set of benefits and drawbacks when it comes to accomplishing their goals.
The majority of the flaws in one strategy are addressed by a different strat- egy. A
successful recommendation engine must deliver both accurate and helpful suggestions,
as well as recommendations that differentiate from those provided by optimization
schemes.
Batuhan Asiroglu [10] argues that today’s product suggestion algorithms re-
quire information about the user’s prior buying habits. This indicates that the
system relies on huge amounts of information and a lot of computational capa-
bilities to develop machine learning methods. Furthermore, these algorithms will
not be able to propose consumer items if they do not even have information on the
user’s prior buying activity. The research team provide a cloth suggestion to a
consumer even without users past buying usage patterns. The researchers offer a
scalable integrated mechanism that uses only a single image to create a textile with
cheap cost and power efficiency.
Zhigao Che [11] describes the creation of an accurate Hadoop-based recommen-
dation engine. Employing crawler technique to gather job recruiting information from
the database, big data technique was utilized to cleanse and aggregate job procurement
data, which could be employed as high degree of influence for pur- suit of its
objectives based on student successes. In practice, it has been demon- strated that this
approach can successfully increase academic employment, assist businesses in
recruiting and selecting talent, enhance the capacity of workforce supervision,
monitoring, and service in educational establishments, as well as pro- vide information
and technology to colleges and universities to help them continue improving their
professional design, among other things.

8
CHAPTER 3

PROJECT STATEMENT

3.1 Problem Definition

To achieve effective product recommendation the paradigm of Natural Language

Processing has been realized through the use of Bag of Words and Term Frequency and
Inverse Document Frequency along with machine learning implementations such as
Fuzzy Artificial Neural Networks and Collaborative filtering to attain mobile
recommendations.

3.2 Proposed Methodology

Figure 3.1: System Diagram

9
CHAPTER 4

SOFTWARE REQUIREMENTS AND

SPECIFICATION

4.1 Introduction

This Software Requirements Specification provides a complete description of

all the functions and constraints of the “Smart Phone Recommendation using
User Reviews and Product Specifications”. The document describes the issues
related to the system and what actions are to be performed by the development team
in order to come up with a better solution.
The recommendation system of today has revolutionized the way we look for things
of interest. This is an information filtering strategy that is used to forecast the user’s
choice. Books, news, articles, music, videos, and movies are among the most prominent
domains where recommender systems are used. In this research, we present a smart
phone recommendation system. To achieve effective product recommendation the
paradigm of Natural Language Processing has been realized through the use of Bag of
Words and Term Frequency and Inverse Document Frequency along with machine
learning implementations such as Fuzzy Artificial Neural Networks and Collaborative
filtering to achieve mobile recommendations.

4.1.1 Project Scope

The purpose of this SRS document is to provide a detailed overview of our soft-
ware product “Smart Phone Recommendation using User Reviews and Product
Specifications”, its parameters and goals. This document describes the project’s
target audience and its user interface, hardware and software requirements. It
defines how our client, team and audience see the product and its functionality.
The main purpose of this system is to enhance the process of mobile
recommendation by using Natural Language Processing along with machine
learning

10
 implementations such as Fuzzy Artificial Neural Networks and Collaborative
filtering.
The scope of this project includes project developer assisted by project guide.
The scope thus far has been the completion of the basic interfaces that will
be used to build the system. The constraints felt thus far by the developer have only
been our weekly story cards, the end-to-end side of the interface, and time to time
brushing on methodology of implementation which schedule the completion of the
project.
The major scope of this project is as follows

The User Interface must be easy to use and should be easily understood by a layman.
The interface must be clean and devoid of any artefacts and errors. The user should
be able to interact with the system with ease and the design should be simple and
coherent.

The proposed system has been deployed with the use of completely free or open
source software. The system does not employ any proprietary software or API’s to
achieve its goals. This is highly useful in the scenario of a crash or an unexpected
error as the system can be redeployed in a very short time due to the open source
nature of the system. All the resources that have been utilized for the development
of the methodology are freesoftware distributions which are very useful and efficient
in addition to being completely open source.

Accurate Smart Phone Recommendation.

4.1.2 User Classes and Characteristics

This project is for the user who is seeking the smart phone recommendation
system. This software is having following characteristics:
End User

Creates user profile

Login in to the system

11
 Update profile if needed

Input user requirements as input

View the recommended product

4.1.3 Assumption and Dependencies

Following are the assumptions:

System is deployed in single machine.

Database is given required permissions.

Dependencies can be:

System speed

IDE behaviour

4.2 FUNCTIONAL REQUIREMENTS

4.2.1 Preprocessing

4.2.1.1 Description and Priority

Input-Dataset

Priority –High

4.2.1.2 Stimulus/Response Sequences

Stimulus: Words

Response sequence: Stopword and stemming.

4.2.1.3 Functional Requirements

Preprocessed list.

12
4.2.2 Bag of Words

4.2.2.1 Description and Priority

Input-Preprocessed list.

Priority –High

4.2.2.2 Stimulus/Response Sequences

Stimulus: Stored words.

Response sequence: Relation factor.

4.2.2.3 Functional Requirements

Bag Word List.

4.2.3 TF-IDF

4.2.3.1 Description and Priority

Input-Feature List

Priority –High

4.2.3.2 Stimulus/Response Sequences

Stimulus: Term Frequency

Response sequence: Inverse Document Frequency

4.2.3.3 Functional Requirements

TF-IDF List.

4.2.4 Fuzzy ANN

4.2.4.1 Description and Priority

Input-Feature List

Priority –High

13
4.2.4.2 Stimulus/Response Sequences

Stimulus: Activation Function

Response sequence: Hidden and Output layer Estimation

4.2.4.3 Functional Requirements

Probability List.

4.2.5 Collaborative Filtering

4.2.5.1 Description and Priority

Input-Probability List

Priority –High

4.2.5.2 Stimulus/Response Sequences

Stimulus: Spatial Distance

Response sequence: Dimension Analysis

4.2.5.3 Functional Requirements

Mobile Recommendation.

4.3 EXTERNAL INTERFACE REQUIREMENTS

4.3.1 User Interfaces

Our system interacts with user on the following occasions:

1. By the user while feeding test Data to the system.

2. By the user while viewing results.

14
4.3.2 Hardware Interfaces

Our system interacts with hardware components on the following occasions:

Our system interacts with the secondary storage device while reading the dataset.

4.3.3 Software Interfaces

• Our system interacts with the Mysql Database server while performing database
operations.

4.3.4 Communication Interfaces

Our system’s different modules are communicating with one another on the fol-
lowing scenarios

1. From Dataset feeding to preprocessing module

2. From preprocessing module to bag of words module

3. From bag of words module to TF-IDF module

4. From TF-IDF module to Fuzzy ANN Module

5. From Fuzzy ANN Module to Collaborative Filtering Module

4.4 Nonfunctional Requirements

4.4.1 Performance Requirements

High Speed: System should process smart phone recommendation system by

using dataset & ratings and Comments. Then system must wait for process
completion.

Accuracy: System should correctly recommend the movie, display the re- sult
accurately. System output should be in user required format. That means all
the intermediate steps need to be display properly.

15
4.4.2 Safety Requirements

The system need to be properly converted into executable file format so thatthe
source code of the system will preserved from any kind of intrusions.

4.4.3 Security Requirements

The system need to be secured using the proper credentials and always the backup
of the system need to be kept to restore the whole system again in worst case scenario.

4.4.4 Software Quality Attributes

Number of functionality provided must be properly working independently.

All logins are secured.

System provides proper way of Fuzzy ANN.

Survivability of the system should be good.

System performs each function accurately.

System is extensible.

System must Evaluate smart phone recommendation.

4.5 System Requirements

4.5.1 Software Requirements (Platform Choice)

1. Platform: JAVA

2. Technology : JDK 1.8

3. IDE: Netbeans 8.2

16
4.3.1 Hardware Requirements

1 Systems of following minimum configuration

Sr. Parameter Minimum Require- Justification

No. ment
1 Processor 2.2 GHz For Fast Processing
2 Hard Disk 200 GB For Fast Processing
3 RAM 4 GB For Fast Processing
4 Monitor, Keyboard 1 Quantity None
and UPS

Table 4.1: Hardware Requirements

4.6 Analysis Model

The Proposed model follows the Waterfall model development cycle for the
development of the source code, as the whole process is working linearly in the
standalone application mode.

Figure 4.1: System Plan

17
4.7 System Implementation Plan

Figure 4.2: Timeline Chart

18
CHAPTER 5

SYSTEM DESIGN

5.1 System Architecture

Figure 5.1: System Architecture

5.2 System Design

Figure 5.2: System Design

19
5.3 Data Flow Diagrams

5.3.1 DFD Level 0

Figure 5.3: DFD Level 0 Diagram

5.3.2 DFD Level 1

Figure 5.4: DFD Level 1 Diagram

20
5.3.3 DFD Level 2

Figure 5.5: DFD Level 2 Diagram

5.4 UML Diagrams

5.4.1 Use Case Diagram

Figure 5.6: Use Case Diagram

21
5.4.2 Activity Diagram

Figure 5.7: Activity Diagram

5.4.3 Sequence Diagram

Figure 5.8: Sequence Diagram

22
5.1.1 Component Diagram

Figure 5.9: Component Diagram

5.4.4 Deployment Diagram

Figure 5.10: Deployment Diagram

23
5.4.5 Package Diagram

Figure 5.11: Package Diagram

5.4.6 State Transition Diagram

Figure 5.12: State Transition Diagram

24
CHAPTER 6

IMPLEMENTATION

6.1 Implementation Detail

Figure 6.1: System Overview

The presented system for sentiment analysis has been depicted in the system
overview given in figure 6.1 above. The sequential steps to achieve the presented
approach have been elaborated in the section below.
Step 1: Data collection and Data feeding – The proposed technique requires the
input of a dataset review for the smartphones. For this purpose a dataset contain- ing
reviews has been extracted from the URL - https://www.kaggle.com/code/moh
itkumar409/smartphone-recommendation-system/notebook#Description:
This dataset consists of a number of different reviews for smartphones and the
requisite information about the reviews. The attributes include the author, country, date,
domain, extract (actual content of the review), language, product, score (rating given
for the product), score max (maximum rating given for the mobile, and source.

25
 The dataset need to be preprocessed before providing it to the system to re-
duce the incidences of any error or redundancy that may impact the performance
negatively. The preprocessing approach has been elaborated in the next step ofthe
approach.
Step 2: Preprocessing – The preprocessing approach is the initial logical step
of the approach that is designed to facilitate the processing of the reviews beforeit
can be provided to the system. The extracted dataset achieved in the previous step
is provided as an input in this step of the procedure. The dataset is in the workbook
format, therefore the JXL library is being used to interface this file with the java code.
The dataset is then converted into a double dimension list that canbe easily processed
by the system.
The process of preprocessing is effective in realization of the conditioning of the
input reviews that are provided. This is highly crucial as the performance of the
execution depends on the effectively of this step of the procedure. The presence of
any unnecessary data can be detrimental to the system as it could result in an error
reducing the accuracy. The redundant data can also take longer to process which can be
problematic to achieve efficiency. The steps for preprocessing are defined below.
Special Symbol Removal – The input string containing the review is provided
to this initial step of the procedure where the special symbols are removed. These
special symbols are used to provide grammatical pauses and other nuances to the text.
This is redundant in our implementation and can be removed without any negative
effects to the review. These symbols, such as ?, , etc are removed from the review.
Stemming – This is the second step in the preprocessing approach where the special
symbol removed review is provided as an input. This step deals with making the
input lightweight which can considerably reduce the processing time of the system.
Majority of the words in the English language come from the same root word and just
differ from one another just by a postfix. These words are then stemmed to their specific
root words, which does not change the meaning of the word. For example, the word
sleeping will be truncated to sleep, which doesn’t

26
change the semantics of the word while making the review lightweight.
Stop Word Removal – The stemmed review is provided to this step of the
preprocessing for the purpose of stop word removal. In the English language, stop
words are words that are used to provide connection between two sentences or combine
two different parts of the same sentence. This is not required by our approach, and the
removal of the stopwords does not provide any penalty in terms of comprehension of
the sentences. Therefore, the words such as and, is, they are eliminated from the review
to achieved the preprocessed review which is provided to the next step for further
processing.
Step 3: Bag of Words – The preprocessed review is provided to this step of
the procedure for effective evaluation of the contents of the review. The content of
the reviews contain a large variety of words which can be Positive, negative or Neutral
in nature. These are useful in evaluation of the overall review and the emotion or the
sentiment of the review. This is one of the most integral parts of the sentiment
analysis module that enables the realization of the text that is being shared in the
reviews.
The bag of words is nothing but a curated dictionary of words that are used to
compare for their presence in the reviews. There are two bag of words or two
dictionaries that are used to achieve the evaluation, namely, the positive bag of words
and negative bag of words. The pseudo protocol used for the assessment determines a
separate list for each of the positive bags and the negative bags are derived from the
assessment of the preprocessed review and utilized further in the proposed system.
Step 4: Term Frequency and Inverse Document Frequency – The preprocessed
review is provided to this module for the realization of one of the central concepts of
Natural Language Processing that is implemented in the proposed system. The Term
Frequency and Inverse Document Frequency determine the amount of time a particular
word is encountered in the review. This is crucial to extract the impor- tance of the
particular word which can be useful in understanding the sentiment of the review.
The equation utilized for the same is provided in the equation 1 given below.

27
 Numberof Documents
TF − IDF = TFXLog Numberof DocumentsContainingWordW
The Inverse Document Frequency provides the frequency of the particular word in
the collection of reviews that can be crucial in achieving the relevance of the word
in the reviews. This data is listed in the form of a double dimension list and
provided to the next step for the purpose of achieving the correlation. The algorithm
for the estimation of the TF-IDF for the reviews is mentioned below in algorithm
1 in algorithm section.
Step 5: Fuzzy Artificial Neural Network – This is one of the most important
steps in the proposed methodology wherein a double dimensional list is utilized to
effectively store the text of the product review as well as the score for correlation of
these values. The fuzzy classification is achieved on this list to effectively segregate the
smartphone review scores and effectively identify the difference between the maximum
and the minimum smartphone review scores. The difference achieved is then
effectively divided into 5 equal parts to achieve effective fuzzy classification labels.
These labels correspond to the fuzzy crisp values that are segregated as very high
medium-low and very low.
The neurons for the artificial neural network are then classified according to
the given rules above depending on the classified product reviews with labels provided
such as excellent satisfactory average disappointed and worst. These neurons are then
subjected to affective evaluation of the hidden layer and the resultant probability
achieved effectively determines the mobile recommendation for the desired level
according to the user.
Step 6: Collaborative Filtering – In this step of the approach, collaborative filtering
is achieved depending on the attributes and mobile recommendations according to those
attributes achieved previously. Another factor in consideration for the collaborative
filtering approach is the effective recommendation achieved by the recommender
system. These two factors are essentially utilized and their cosine similarity is
effectively determined but converted into an effective vector for the cosine angle
analysis. This allows for improvement in the recommendation systems and significantly
improves the execution performance and accuracy of the mobile recommendation
considerably. The resultant collaborative filtering index

28
utilized effectively extract the nearest coefficient in the form of the smartphone which
is provided as a recommendation to the user. The Collaborative Filtering step is
detailed in algorithm 2 in algorithm section.

6.2 Algorithm Details

Algorithm 1 TF-IDF Estimation

1: Start
2: Read the Preprocessed Reviews string
3: divide string into words on space and store in a vector V
4: FOR i =0 to N (Where N is the length of V)
5: W= V[i]
6: Count W for the respective Reviews as TF
7: Count W for the all other Reviews that is DF
8: IDF= log ( DF)
9: TF-IDF= TF* IDF
10: END FOR
11: Stop

Algorithm 2 Collaborative Filtering

//Input: PreviouslyrecommendedmobilePR, CurrentRecommendationCR
//Output: Collaborative Filtering index List - CFI

1: Start
2: For i=0 to Size of CR
3: CRL of CRLi (CRL Current Recommendation List)
4: IF CRL ∈ PR
5: THEN — CRL – PR — < CFL [CFL - Collaborative Filtering Limit variable]
6: Set Collaborative Filtering index CFI
7: Create CFI List
8: END For
9: Return CFI
10: Stop

6.3 Tools and Technologies Used

The presented approach for the purpose of achieving mobile recommendation has been
realized through the use of the java programming language. The NetBeans IDE is
being used to achieve the development of the system. The development

29
machine is powered by 8GB of RAM assisted by 1TB of storage and an Intel Core i5
processor. The database responsibilities and requirements are handled by the MySQL
Database server.

30
 CHAPTER 7

SOFTWARE TESTING

7.1 Type of Testing Used

7.1.1 Performance Testing

The performance of the system is measured by the Precision and Recall for
the given dataset input

7.1.2 System Testing

Checked the steadiness of the space provided by the system for the maximum
number of rows in the dataset. This is done by feeding the maximum number of
data rows to learn and so that set the threshold by the Virtual machine option of the
Netbeans IDE.

7.1.3 Recovery Testing

Our system can be recovered in span of 2 hour after crashing. Within two hour we
can install all the Software and deploy our system to work as in the past.

7.1.4 Security Testing

7.1.4.1 Stress Testing

The System is well equipped to stand against the breakdown point of maximum
number of dataset rows decided by the IDE’s virtual machine settings, beyond that the
memory overflow exception may arise.

7.1.4.2 Unit Testing

All the modules are independently handled developed and ran to get proper output
and finally they are integrated to get the whole output.

31
7.1.4.3 Black Box Testing

Compatibility analysis is done by passing the output of one module to another and
checking for estimated output of Mobile Recommendation.

7.1.5 Integration Testing

When all the individual modules are integrated into our system, that yields a proper
system of Mobile Recommendation system that is cross examined for the desired
output.

7.2 Test Cases

Test Cases for User

ID TEST CASE USER INPUT PASS CRITERIA

U REG User Registra- User Attributes All the user data is
tion stored in the database
successfully.
U LOGIN User Login User’s Username and Successful login when
Login valid credentials en-
tered.
U Update User Profile Up- Updated Attributes All the user attributes
date are successfully up-
dated in the database
U Search User Searching User Search At- User get successful
For mobile tributes Recommendation from
Rec- the System
ommendation
U History User View His- Username User Successfully
tory View the History of
the Recommendation

Table 7.1: Test Cases for User

32
CHAPTER 8

RESULTS

8.1 Snapshot

Figure 8.1: Login Frame

Figure 8.2: Registration Frame

33
Figure 8.3: User Operation Frame

Figure 8.4: Edit Profile Frame

34
 Figure 8.5: Search Frame

Figure 8.6: Output Frame

Figure 8.7: View History Frame

35
8.2 Results and Discussion

The presented approach for the purpose of achieving mobile recommenda- tion
has been realized through the use of the java programming language. The NetBeans
IDE is being used to achieve the development of the system. The de- velopment
machine is powered by 8GB of RAM assisted by 1TB of storage and an Intel
Core i5 processor. The database responsibilities and requirements are handled by
the MySQL Database server.
The evaluation of the presented approach has been performed to measure the
performance of the methodology. The performance evaluation is necessary to determine
the proper deployment of the various modules. For the assessment of the approach
the paradigm of Precision and Recall has been used.
Performance Evaluation based on Precision and Recall
The evaluation of the Precision and Recall has been initiated to determine
the accuracy of the system performance for mobile recommendation. The system
utilizes the review dataset for the smartphones as an input. The evaluation of the
recommendation has been performed through the combination of several modules, such
as Bag of Words, TF-IDF along with Fuzzy ANN and Collaborative filtering.
These modules must be implemented accurately and the output for the same must be
accurate from one module and another to achieve the accurate recommendation. The
performance for the analysis of the recommendation is performed through the use of
Precision and Recall metric.
Precision and Recall are extremely useful metrics that determine the execution
accuracy of the system. The precision metric is defined as the absolute accuracy of the
system and on the other hand the recall is considered as the relative accuracy of the
system.
For the evaluation purposes the precision as the ratio of number of accurate mobile
recommendations by the total number of reviews extracted. The recall metric is
considered as the ratio of accurate mobile recommendation, by the total number of
inaccurate mobile recommendations.
The Precision and Recall metrics have been mathematically illustrated below.

36
 A = The number of accurate mobile recommendations for the given reviews

B= The number of inaccurate mobile recommendations for the given reviews

C = The number of accurate recommendations that are not done for the given
reviews

So, precision can be defined as

Precision = (X / (X+Y))
*100 Recall = (X / (X+ Z))
*100
The experimental outcomes have been illustrated in a tabular format in thefigure
8.8 given below. These values have also been depicted graphically in the line
graph provided in the figure 8.9 below.

Figure 8.8: : Precision and Recall Measurement for the Mobile recommendation

Figure 8.9: Comparison of Precision and Recall for the Mobile recommendation

37
 It can be noticed from the graphs that the performance indicated through the
values of Precision and Recall is quite good for the mobile recommendation achieved
through the use of Fuzzy ANN and Collaborative Filtering. The perfor- mance of the
system takes the smartphone reviews as an input and achieves highly satisfactory
performance. The precision and recall values of 88.25 and 90.08 are better than
expected for the first time implementation of such a system for the recommendation.

38
CHAPTER 9

CONCLUSION

9.1 Conclusion

The presented methodology for Mobile recommendation on the product re-

views dataset and user requirements through the use of natural language process- ing
along with Fuzzy Ann and Collaborative Filtering has been outlined in this
research. The presented system takes the user requirement as well as user reviews as
input, the reviews are in the form of an excel file which is first preprocessed. The
preprocessing involves the removal of special symbols from the reviews along with
stemming procedures and stop word removal before giving it as an input to the
system. The preprocessed review is subjected to bag of words evaluation which
compares the review for the presence of positive and negative words. These words are
counted and the next type of the procedure is performed for the determination of the
term frequency and inverse document frequency. The TF-IDF approach is useful as
it allows for effective evaluation of the importance of the words in the review
which are then used as an input to the Fuzzy ANN. The Fuzzy ANN takes the
input and generates the neurons that are then processed using fuzzy crisp val- ues to
achieve the output layer values. This parameters are then provided to the
collaborative filtering approach which classifies the reviews into neutral reviews
for positive and negative reviews. The experimental evaluation has depicted that
the presented technique has achieved satisfactory levels of precision and recall.

9.2 Future Scope

The future research prospects can be focused on achieving this mobile

recommendation approach into to an API for universal application and integration.

39
APPEX A

REFERENCES

[1] E. Shibamoto, C. Kittirojrattana, C. Koopipat, A. Hansuebsai and K.

Takano, ”A Recommendation System of Sightseeing Places based on User’s
Behavior of Taking and Editing Photos,” 2019 IEEE Pacific Rim
Conference on Com- munications, Computers and Signal Processing
(PACRIM), 2019, pp. 1-6, doi: 10.1109/PACRIM47961.2019.8985118.

[2] F. Xie, Z. Cao, Y. Xu, L. Chen and Z. Zheng, ”Graph Neural Network and
Multi-view Learning Based Mobile Application Recommendation in Hetero-
geneous Graphs,” 2020 IEEE International Conference on Services Comput-
ing (SCC), 2020, pp. 100-107, doi: 10.1109/SCC49832.2020.00022.

[3] R. Liu et al., ”Understanding Mobile Users’ Privacy Expectations: A

Recomm endation-Based Method Through Crowdsourcing,” in IEEE
Transactions on Services Computing, vol. 12, no. 2, pp. 304-318, 1
March-April 2019, doi: 10.1109/TSC. 2016.2636285.

[4] D. I. Sumiko and A. A. Agus, ”The Effect of Product Recommendation in

Youtube to Consumer Impulsive Buying of Smartphone Product,” 2020 3rd
International Conference on Computer and Informatics Engineering (IC2IE),
2020, pp. 206-211, doi: 10.1109/IC2IE50715.2020.9274622.

[5] J. Liu, Y. Qiu, Z. Ma and Z. Wu, ”Autoencoder based API Recommen-

dation System for Android Programming,” 2019 14th International Confer-
ence on Computer Science & Education (ICCSE), 2019, pp. 273-277, doi:
10.1109/ICCSE.2019.8845349.

[6] P. Rustgi and C. Fung, ”Demo: DroidNet - An Android Permission Con-

trol Recommendation System Based on Crowdsourcing,” 2019 IFIP/IEE

40
 Symposium on Integrated Network and Service Management (IM), 2019, pp. 737-
738.

[7] Y. Kato and K. Yamamoto, ”Development of Sightseeing Spot Recommen-

dation System Considering Users’ Visit frequency,” 2020 Joint 11th Interna-
tional Conference on Soft Computing and Intelligent Systems and 21st In-
ternational Symposium on Advanced Intelligent Systems (SCIS-ISIS), 2020,
pp. 1-8, doi: 10.1109/SCISISIS50064.2020.9322776.

[8] N. Kadri, A. Ellouze and M. Ksantini, ”Recommendation system for hu-

man physical activities using smartphones,” 2020 2nd International Confer-
ence on Computer and Information Sciences (ICCIS), 2020, pp. 1-4, doi:
10.1109/ICCIS49240.2020.9257671.

[9] S. Roy, M. Sharma and S. K. Singh, ”Movie Recommendation System Us-

ing Semi-Supervised Learning,” 2019 Global Conference for Advancement in
Technology (GCAT), 2019, pp. 1-5, doi:
10.1109/GCAT47503.2019.8978353.

[10] B. AŞIROĞLU, M. İ . ATALAY, A. BALKAYA, E. TÜZÜ

NKAN, M. Dağtekin and T. ENSARI˙, ”Smart Clothing Recommendation
System with Deep Learn- ing,” 2019 3rd International Symposium on
Multidisciplinary Studies and In- novative Technologies (ISMSIT), 2019,
pp. 1-4, doi: 10.1109/ISMSIT.2019. 8932738.

[11] Z. Chen, W. Liang, X. Gao, Z. Zhou and M. Wu, ”Research on the Ac-
curate Recommendation Management System for Employment of College
Graduates on Hadoop,” 2019 5th International Conference on Big Data
and Information Analytics (BigDIA), 2019, pp. 19-22, doi: 10.1109/Big-
DIA.2019.8802855.

41