Product Recommendation System

SUBMITTED BY
Group Number :- 21

Team Members:

1) Name:- Akash Kumar

Enrollment No.:- 1911100001090
Reg. No. :-190000906301

2) Name:- Manash Kumar Singh

Enrollment No. :-1911100001103
Reg. No. :-190001039829

3) Name:- Surya Poddar

Enrollment No. :- 1911100001076
Reg. No. :- 190000763985

4) Name :- Altaf Ali

Enrollment No. :- 1911100001048
Reg. No. :- 190000485105
 Under the Supervision of

Mr. Anirban Mitra

Head of Department
Dept. of Computer Science and Engineering

Signature & Date (Supervisor)

 Acknowledgement

We would like to convey my heartfelt gratitude to Mr. Anirban Mitra for her
tremendous support and assistance in the completion of my project. I would also
like to thank our HOD , Mr. Anirban Mitra, for providing me with this wonderful
opportunity to work on a project with the topic Product Recommendation
Service.
The completion of the project would not have been possible without theirhelp and
insights.
 Certificate from the Head of the Department of the University

This is to certify that Mr/Ms ............................................... is a final semester/year student of .............................

(B.Tech/M.Tech etc) in
....................................... in the Sister Nivedita University. He/She will appear/has appeared for the final year
examination
during
................... 2023. His/Her result is awaited. He/She has obtained a cumulative percentage/CGPA of . in the
previous semesters/years exams of this degree.

Place: Signature and Seal of

Head of the Department of Computer Science and Engineering
 Content Page No
1) Introduction ................................................................................................................................ 4-12

i. Industrial Application
ii. Recommendation Systems And Types

2) Methodology ................................................................................................................................13-15

i. The Problem
ii. Need For Study

3) Literature Review ........................................................................................................................ 16-20

i. Techniques and their Limitations

4) Detailed Methodology ................................................................................................................. 21-26

5)Conclusion and Future Scope.....................................................................................................27-29

6) References ................................................................................................................................... 30-32

 INTRODUCTION
To enhance the customer experience and to boost up the sales of products, almost all of the companies are trying to make
some sort of mechanism that is nothing but a recommendation systems. So to finalize this product recommendation system
comes into the light.

It is evident that the pace that technology advances have been increased over the last decades. Scientific discoveries and
technological growth introduced to people a huge variety of options and possibilities. One of the most important
advantages that technology offers is the direct and easy access to information. Nowadays access to vast networks of
information is easy and people can be informed about almost anything they desire.

Even though ease of access provided people with the ability to acquire the needed information, they are now facing a new
obstacle: this of easily finding what they need. On one hand, information abundance covers the majority of needs but on
the other hinders accessibility to information truly valuable to the user. The term that describes this phenomenon is
“Information Overload”. Often users are presented with seemingly similar information to their inquiry but irrelevant to
their actual needs, rendering this way the discovery of the desiredknowledge a difficult task.

Recommender systems (RSs) fall in the category of information filtering approaches. The term describes the software
tools and techniques that are utilized in order to recommend items of interest and value to the user.
 We have chosen this topic as in today's technical era, every startup or a company attempt to establish a better sort of
communication between their products and the users, and for that purpose, they require a type of mechanism which can
promote their product effectively, and here the recommender system serves this motive. These systems are proved to be very
beneficial in variety of domains involving music, books, movies, researcharticles and products in common.

As we can see today many e-commerce apps and websites are using product recommendation systems to
make it user friendly as like:

a. Amazon uses Recommendation System as a marketing tool all over its website and app.

b. Myntra uses Recommendation System as a marketing tool all over its website and app.

c. Netflix uses an AI Recommendation system for movie streaming and searches.

d. Best Buy is the company which uses the recommendation system to increase revenues and upgrade
customer experience.
 ● Industrial Application
The importance of Recommendation Systems can easily be seen by their various existing implementations in well-known
websites and apps.

Amazon:

Amazon created its recommendation system in 1998 implementing an item-based collaborative filtering algorithm.
Based on users’ past behavior, context and ratings it manages to offer a unique customer experience.

The system is embedded and utilized in multiple ways throughout the whole shopping experience. It starts even from
the beginning when a number of products is recommended based on previously seen items. It recommends products
that were bought together with the product being watched. Furthermore it recommends items that might be related to
the product in order to discover unknown interests.

YouTube:

YouTube also has developed its own recommendation engine. YouTube’s RS is a top-N recommender and it aims to
provide personalized recommendations based on user recent behavior. More over the second goal of the system is to
promote through recommendations the wide range of available content that offers. Some of the main challenges that the
system had to face was the lack of meta-data associated with the videos while also the short user interactions that made
the discovery of user intent a difficult task.
Lastly YouTube considers the way the recommendations are being presented to its users as highly important. It offers
explanations as to why they are recommended and advances personalization even more by allowing its users to control
where and how many recommendations should appear.

Netflix:

Netflix does not create recommendations based on a single algorithm, but rather utilizes a number of different
ones regarding the use case they were designed for. Taking into account the fact that the longer time a user
spends in searching for a show the more likely it is for him to stop using the service. Company focused its system
in being able to provide suggestions that will draw the attention of the user in the top of the list of shows. On
top of that it should be made clear to the user why each show is being recommended.

Netflix Prize 2009 offered 1 million dollars to the team with the higher prediction accuracy algorithm. As a result
many algorithms are being used in production even today.

As already noted the system is composed by a number of algorithms. Personalized Video Ranker (PVR) is used to
offer personalization and is responsible to define the order in which videos of a specific genre appear. Essentially
Netflix also uses different algorithms for ordering shows in the rows of Trending rows and Continues watching.
Moreover Video-Video similarity is used to suggest show in a “Because You Watched” section.
 • Recommendation Systems and Types:

A recommender system is used by every e-commerce

website these days like Flipkart, Amazon, Netflix and
YouTube uses such technology to gain more
engagement by the users. There are many different
types of recommender system that are used at
different platforms which are build according to their
architecture.
Different types of recommender systems are discussed
below:
I. Collaborative-filtering Recommendation
System

This approach is typically used to capture and

analyze user's actions, interests or preferences
information and determine what they want
based on their similarities to other users . One
important benefit of a collective filtering
approach is the ability to reliably suggest
complex objects like films without the need of
"understanding" the product itself. This approach
does not focus on computer analyzable
information. Collaborative filters rely on the
premise that parties who have agreed in the past
agree in the future and prefer related things in
the past. For example, if a person A likes item 1,
2, 3 and B, it's equivalent to 2,3,4; and A would
like item 4 and B would like item 1. Collaborative
filtering can be further divided into 2 categories:
 User-Based Collaborative Filtering:

Model is looking for lookalike buyers and offering

items focused on the choices made by its lookalike.
This is a highly efficient and at the same time
consuming algorithm. This method of filtering includes
any pair of consumer details that takes time to be
computed. This algorithm is difficult to implement for
large base platforms.

 Item-Based Collaborative Filtering:

It is similar to the above algorithm, but the Model aims

to look at the item in the same way instead of finding a
consumer look. After the model has a matrix that looks
the same, it can conveniently advise a customer who
has bought every object in the shop. This algorithm
takes far less resources than collaborative filtering
between user and user. The algorithm takes much less
time for a new customer than it does for the user-user
collaborative filtering as all similarity values among
customers are not necessary. In its recommendation
engine, Amazon uses this strategy to present similar
goods that enhance sales.
II. Content-Based Recommendation system :

The system learns to make suggestions based on the

previous responses submitted by the user by evaluating
the feature similarities between the items. The user
profile based on historical data of previously rated
items is generated by a Content based
recommendation system . A user profile reflects the
interests of the user and is also able to adapt to new
interests. User profile is matched with the content
object's features and is essentially the mechanism of
recommendation. A decision indicating the user's
interest in the item results from this procedure.
III. Hybrid Recommendation System:

New studies have shown that it can be more efficient to

combine collective and content-based recommendations.
Hybrid methods can be applied by independently making and
then integrating content-based and collaborative predictions.
In addition, it adds content based capabilities to a collaborative
approach and vice-versa; or it integrates methods into a single
paradigm. Several trials have been carried out to compare the
hybrid's success with pure content-based and collaborative
techniques. They show that hybrid methods can be more
precise than pure approaches. Such approaches may be used
to solve typical challenges, such as cold start or data paucity, in
recommendation systems. Netflix's use of hybrid
recommendation systems is a prime example. The website
recommends by comparing the viewing and browsing
behaviors (i.e. collaborative filtering) of similar users as well as
by offering movies which share features with highly rated
movies (content-based filtering).
 METHODOLOGY
• The Problem
Recommender systems are proved to be a valuable feature which enhances user experience as well as an important tool in the
hands of the service provider. E-commerce sites are a domain that could greatly benefit from the advantages a Product
Recommendation System has to offer. These websites though must take into consideration all the factors that affect user
experience in order to increase revenue. Users often use e-commerce systems because they can do their shopping with comfort
and ease. It is evident that the design of an ecommerce website should be done in such a way that it does not harass users with
constant questions about their preferences and force them to provide feedback.

Additionally the system needs to make the whole process of shopping easy and fast, otherwise it risks losing sales and loyalty. It
obvious that these implementations suffer from the lack of explicit input and feedback from the users. Even though recommender
systems have been extensively researched, most of the times research focused on use cases where the user provided input
explicitly via a ranking system. There is still more room to design and enhance recommendation algorithms that are based solely on
implicit information gathered by monitoring user behavior. This way ecommerce systems can offer personalized recommendations
without deteriorating the experience of its users. Such implicit information, even though they are easily gathered from monitoring
users browsing behavior and exploiting his purchase history, often concern a small number of items rather than the whole set
offered by the provider. Thus, an additional challenge that should be addressed is data sparsity that these implementations suffer
from.
Some of the problems which we understand can be solved are :

(a) Cold Start Problem: A cold start problem is a problem that arises when no information is found about the user or item
in the systems. Collaborative filtering recommender system which needs mandatory information about user and item
before recommending fails.

(b) Sparsity Problem: In the recommendation system, it has been noticed that most users use the system butdo not give
rating for feedback to the system in a proper way. So even though we may have many users using the recommender
system various Times, it is possible that we have very few ratings from those users about different items which they have
liked or purchased or even disliked.

(c) Scalability: When the system faces scalability issues it becomes slow it starts feeling it start giving
problems which it has never given when a load of users recommendation were less.

(d) Over Specialization problem: This problem arises if the recommended items are too similar to each other. On
seeing same item again and again and not any unique suggestion may lead user to switch the shopping website whose
recommender system offers better, interesting and personalized results for the user.

(e) Lack of Data: Perhaps the biggest issue facing recommender systems is that they need a lot of data to effectively
make recommendations.
• Need For Study:

The purpose of this study is to examine implementations of various recommendation

algorithms. It aims at developing a Product Recommendation Engine suitable for e- commerce
websites, thus it will be based solely on implicit input. As input we use a combination of implicit
sources. This combination consists of browsing user behavior, add to cart actions as well as
their purchase history. The implemented algorithm is a Matrix Factorization algorithm which
incorporates confidence and it based on the work. Additionally, the data sparsity problem is
being addressed by utilizing association rule mining.
 LITERATURE REVIEW

1. In 2018, Li et al.[1] have developed a recommendation algorithm using Weighted Linear Regression Models
(WLRRS). The proposed system was put to experiment using the Movie Lens dataset and it presented better
classification and predictive accuracy.

2. In 2019, Hammou et al. [2] proposed a Big Data recommendation algorithm capable of handling large scale
data. The system employed random forest and matrix factorization through a data partitioning scheme. It was
then used for generating recommendations based on user rating and preference for each item. The proposed
system outperformed existing systems in terms of accuracy and speed.

3. In 2016, Liao et al. [3] proposed a recommender model for online customers using a rough set association
rule. The model computed the probable behavioral variations of online consumers and provided product
category recommendations for e-commerce platforms.
4. In 2020, Ullah et al. [4] have implemented an image-based service recommendation model for online
shopping based random forest and Convolutional Neural Networks (CNN). The model used JPEG
coefficients to achieve an accurate prediction rate.

5. In 2020, Esteban et al. [5] have implemented a hybrid multi-criteria recommendation system concerned
with students’ academic performance, personal interests, and course selection. The system was
developed using a Genetic Algorithm (GA) and aimed at helping university students. It combined both
course information and student information for increasing system performance and the reliability of the
recommendations.

6. In 2021, Dhelim et al. [6] have developed a personality-based product recommending model using the
techniques of meta path discovery and user interest mining. This model showed better results when
compared to session-based and deep learning models.
Techniques and their Limitations:
1. Collaborative Filtering:

Collaborative filtering technique is the most mature and the most commonly implemented.
Collaborative filtering recommends items by identifying other users with similar taste; it uses their
opinion to recommend items to the active user,[7].
The system uses collaborative filtering method to overcome scalability issue by generating a table of
similar items offline through the use of item-to-item matrix. The system then recommends other
products which are similar online according to the users’ purchase history.

• Limitations:

Collaborative approaches exhibit cold-start, synonymy, sparsity and scalability problems.

These problems usually reduce the quality of recommendations.
 2. Content-Based Filtering:
Content-based filtering techniques normally base their predictions on user’s information, and they
ignore contributions from other users as with the case of collaborative techniques.
Fab relies heavily on the ratings of different users in order to create a training set and it is
an example of content-based recommender system.
The system makes use of a user interface that assists users in browsing the Internet; it is able to
track the browsing pattern of a user to predict the pages that they may be interested in [7].

• Limitations:

Content based filtering techniques are dependent on items’ metadata. That is, they require rich description
of items and very well organized user profile before recommendation can be made to users. This is
called limited content analysis. So, the effectiveness of CBF depends on the availability of descriptive data.

Content overspecialization is another serious problem of CBF technique. Users are restricted to
getting recommendations similar to items already defined in their profiles.
3. Hybrid filtering :

Hybrid filtering technique combines different recommendation techniques in order to gain better
system optimization to avoid some limitations and problems of pure recommendation systems.
The idea behind hybrid techniques is that a combination of algorithms will provide more
accurate and effective recommendations than a single algorithm as the disadvantages of one
algorithm can be overcome by another algorithm. Using multiple recommendation techniques
can suppress the weaknesses of an individual technique in a combined model. The
combination of approaches can be done in any of the following ways: separate implementation
of algorithms and combining the result, utilizing some content-based filtering in collaborative
approach, utilizing some collaborative filtering in content-based approach, creating a unified
recommendation system that brings together both approaches.

They can be classified based on their operations into weighted hybrid, mixed hybrid, switching
hybrid, feature-combination hybrid, cascade hybrid, feature-augmented hybrid and meta-level
hybrid .
DETAILED METHODOLGY:
Collaborative Filtering:

Collaborative filtering algorithms can be divided in categories based on their design. The most known distinction is
item-based and user-based filtering.

 User-Based Collaborative Filtering:

Common collaborative algorithms used user similarity for their predictions. Similar users are usually described as
neighbors. In user-based algorithms this association is utilized in order to predict preference of a user towards a specific
item. This is achieved by processing the ratings of all the user’s neighbors. A very simple form of this approach could
take as prediction the average rating score of all the neighbors to this item.

 Item-Based Collaborative Filtering:

In a similar manner to user-based algorithms, item-based collaborative filtering algorithms base their
recommendations considering similarities, alas in this case we compare item similarity instead of user. To make it
clearer in order to predict a user’s rating for a specific item we 18 take into account its similar items and the ratings the
user have provided for them.
Content-based Filtering systems:

It aims to create profiles of user interests. These profiles can be

constructed from various sources. They could consist of explicitly stated preferences using
predefined terms or categories. In addition they can be constructed with the use of implicit
information from past interactions like queries, views or time spent browsing.

Algorithm used in Content-Based Filtration is given below:

 Classification Algorithms
This algorithms are used extensively in content based algorithms since they learn a model of
user’s interests. The classifier given a model and a new item can predict if the specific item is of
interest to the user.
Some approaches use probabilistic models on past user interaction. Naïve Bayes used for text
classification is an accurate and popular example of such algorithms. The idea of the algorithm is
to calculate the probability of a document to belong to a specific class.

 Relevance feedback and Rocchio’s Algorithm

Another established algorithm used in content-based filtering, and more specifically in text categorization domains
is Relevance feedback and Rocchio’s Algorithm. The main idea is to let users specify if the items returned by the
system are interesting to them. The feedback can then be used to improve the retrieval of items. Relevance
feedback has been the main focus of many researchers that aimed to develop information retrieval algorithms.
Rocchio’s algorithm is a relevance feedback algorithm. The algorithm modifies the initial query by weighting relevant
and irrelevant documents. The approach as described in creates two document prototypes.
 Block Diagram of Content-Based Filtering
Flow Chart of Content-Based Filtering
Hybrid filtering:

Hybrid filtering technique combines different recommendation techniques in order to gain better system
optimization to avoid some limitations and problems of pure recommendation systems. The idea behind hybrid
techniques is that a combination of algorithms will provide more accurate and effective recommendations than
a single algorithm

 Hybrid algorithms

It consists of various combinations of the previous approaches. Hybrid

approaches usually are more efficient and more accurate because they combine techniques to
alleviate limitations. Though they are not suitable for simpler cases because of the increased
complexity level they present during the design and implementation phase.
Success of a recommender system is based in a number of properties. Context of the specific
domain where the recommender will be deployed heavily affects the overall performance of the
system. Different use cases of recommender systems aim to satisfy different needs. The
properties that affect the systems success have to be identified for evaluating the system.
Most recommendation systems implement a prediction algorithm, thus prediction accuracy
metrics are commonly used for system evaluation. Root Mean Squared Error (RMSE) is
commonly used to measure accuracy.
 Block Diagram of Hybrid Filtration
Flow Chart Of Hybrid Filtration
 Conclusion

Recommender systems have attracted the attention of researchers and academicians. In this paper,
we have identified and prudently reviewed research papers on recommender systems focusing on
diverse applications, which were published between 2016 and 2021. This review has gathered diverse
details like different application fields, techniques used, simulation tools used, diverse applications
focused, performance metrics, datasets used, system features, and challenges of different
recommender systems. Further, the research gaps and challenges were put forward to explore the
future research perspective on recommender systems. Overall, this paper provides a comprehensive
understanding of the trend of recommender systems-related research and to provides researchers
with insight and future direction on recommender systems. The results of this study have several
practical and significant implications:

• Based on the recent-past publication rates, we feel that the research of recommender systems will
significantly grow in the future.

• A large number of research papers were identified in movie recommendations, whereas health,
tourism and education-related recommender systems were identified in very few numbers. This is
due to the availability of movie datasets in the public domain. Therefore, it is necessary to develop
datasets in other fields also.
• There is no standard measure to compute the performance of recommender systems. Among 60
papers, 21 used recall, 10 used MAE, 25 used precision, 18 used F1-measure, 19 used accuracy
and only 7 used RMSE to calculate system performance. Very few systems were found to excel
in two or more matrices.

• Java and Python (with a combined contribution of 27%) are the most common programming
languages used to develop recommender systems. This is due to the availability of a large
number of standard java and python libraries which aid in the development process.

• Recently a large number of hybrid and optimizations techniques are being proposed for
recommender systems. The performance of a recommender system can be greatly improved by
applying optimization techniques.

• There is a large scope of research in using neural networks and deep learning-based methods
for developing recommender systems. Systems developed using these methods are found to
achieve high-performance accuracy.
 Future Scope

This research will provide a guideline for future research in the domain of recommender systems.
However, this research has some limitations. Firstly, due to the limited amount of manpower and
time, we have only reviewed papers published in journals focusing on computer science,
management and medicine. Secondly, we have reviewed only English papers. New research may
extend this study to cover other journals and non-English papers. Finally, this review was conducted
based on a search on only six descriptors: “Recommender systems”, “Recommendation systems”,
“Movie Recommend*”, “Music Recommend*”, “Personalized Recommend*” and “Hybrid
Recommend*”. Research papers that did not include these keywords were not considered. Future
research can include adding some additional descriptors and keywords for searching. This will allow
extending the research to cover more diverse articles on recommender systems.
 References:
1. Li C, Wang Z, Cao S, He L. WLRRS: A new recommendation system based on weighted linear regression models. Comput
Electr Eng. 2018;66:40–7. https://www.sciencedirect.com/science/article/abs/pii/S0045790617328124

2. Hammou BA, Lahcen AA, Mouline S. An effective distributed predictive model with matrix factorization and
random forest for big data recommendation systems. Expert Syst Appl. 2019;137:253–65.
https://www.sciencedirect.com/science/article/abs/pii/S0957417419304464

3. Liao SH, Chang HK. A rough set-based association rule approach for a recommendation system for online
consumers. Inf Process Manage. 2016;52:1142–60.
https://www.sciencedirect.com/science/article/abs/pii/S0306457316301224

4. Ullah F, Zhang B, Khan RU. Image-based service recommendation system: A JPEG-coefficient RFs approach.
IEEE Access. 2020;8:3308–18. https://ieeexplore.ieee.org/abstract/document/8943224

5. Esteban A, Zafra A, Romero C. Helping university students to choose elective courses by using a hybrid multi-
criteria recommendation system with genetic optimization. Knowledge-Based Syst. 2020;194:105385.
https://www.sciencedirect.com/science/article/abs/pii/S0950705119306306

6. Dhelim S, Ning H, Aung N, Huang R, Ma J. Personality-aware product recommendation system based on user
interests mining and metapath discovery. IEEE Trans Comput Soc Syst. 2021;8:86–98.
https://ieeexplore.ieee.org/abstract/document/9269396
7. F.O. Isinkaye, Y.O. Folajimi, B.A. Ojokoh,Recommendation systems: Principles, methods and evaluation,Egyptian
Informatics Journal,Volume 16, Issue 3,2015,Pages 261-273, ISSN 1110-8665,
https://www.sciencedirect.com/science/article/pii/S1110866515000341#b0080

8. Roy, D., Dutta, M. A systematic review and research perspective on recommender systems. J Big Data 9, 59 (2022).
Published03 May 2022. https://journalofbigdata.springeropen.com/articles/10.1186/s40537-022-00592-
5#citeas

9. B. Song and J. Gao, “The enhancement and application of collaborative filtering in e-learning system,”
in Proceedings of the International Conference in Swarm Intelligence, pp. 188–195, Springer, Berlin,
Germany, 2014. https://link.springer.com/chapter/10.1007/978-3-319-11897-0_22

10. D. Godoy and A. Corbellini, “Folksonomy-based recommender systems: a state-of-the-art

review,” International Journal of Intelligent Systems, vol. 31, no. 4, pp. 314–346, 2016.
https://onlinelibrary.wiley.com/doi/abs/10.1002/int.21753

11. Z. Zhu, X. You, P. C. L. Chen et al., “An adaptive hybrid pattern for noise-robust texture analysis,” Pattern
Recognition, vol. 48, no. 8, pp. 2592–2608, 2015.
https://www.sciencedirect.com/science/article/abs/pii/S0031320315000023

12. Y. Yang and S. Hu, “Exploring reviews and ratings on reviews for personalized search,” in Proceedings of
the 14th ICWL Conference on Web-Based Learning, pp. 252–263, Springer Press, Hong Kong, China,
2015. https://link.springer.com/chapter/10.1007/978-3-319-32865-2_16