Volume 4, Issue 2, February – 2019 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Movie Recommendation System using Simple

Recommender-Based Approach
Sushmita Roy
(Author), Post-Graduation Student
Department of Information Technology (IT)
Thakur College of Science and Commerce (TCSC)
Mumbai, India

Abstract:- With the expeditious development of their interest in instances like - Job posts, Online shopping,
Information and Communications Technology and Movie of Interest, Suggested Video, Games of Interest,
Web technologies, an enormous information is Facebook friend that you know and People who have
currently available and this results in a condition which bought this have also bought this etc. have now become
in some cases is known as information overload. Due to easier.
these circumstances, it’s growing arduous for a person
to discover and to access information for taking II. WHAT IS RECOMMENDATION SYSTEM OR
decisions expeditiously to arrive at an effective RECOMMENDATION ENGINE ?
conclusion. To perorate this nut, there are filtering
systems for information, such as the recommendation Recommendation Systems, often termed as
system or recommendation engine, considered here in Recommender Systems or Recommendation Engines, are
this paper, that help a person in identifying significant, simply the usage of various algorithms whose aim is to
tectonic and possible services or products of interest provide relevant and accurate products to customers by
based on the preferences given by him/her. Several means of filtering of useful data from a large pool of
approaches exist like - simple recommendation, information base. Recommendation systems discover
popularity-based recommendation, collaborative patterns in the data set by learning and analyzing user
filtering, content-based filtering demographic-based choices and behavior and produce the outcomes that best
filtering and keyword or metadata-based filtering. This relates to the interests of user.
study has been undertaken to investigate the simple
recommender-based approach. The efficiency of this Thus, recommendation systems predicts the
proposed methodology is verified by experiments based “preference” or “rating” that a customer would give to a
on The Movies Dataset. product.

Keywords:- Recommendation System; Recommendation Figure 1 shows a typical usage of recommendation

Engine; Movie Recommendation System. system.

I. INTRODUCTION

Quite often we might think which algorithm IT giant

like Google would be using to maximize its target ad
revenue or often we wonder how e-commerce websites lure
us to buy products through their choices such as “people
who have bought product X also have bought Product Y” or
“customers who have bought this have also bought this” or
how can Facebook suggest us friends or suggest to tag
friend in a group of pictures.

The answer to all of these is using a recommender

system or recommendation engine. Now, with the large
amount of information available on the Web and with the
significant increase in the number of people, it is growing
important for organizations to search and provide with the
relevant information according to their preferences and
tastes.

Currently organizations are involved into building

recommendation systems or recommendation engine by Fig 1:- Recommendation System Usage
learning and analyzing the past behavior of the customers.
Hence, giving users with recommendations and choices of

IJISRT19FB208 www.ijisrt.com 444

 Volume 4, Issue 2, February – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
III. RECOMMENDATION SYSTEMS – THE NEED  Music
OF THE COMMERCE ERA  Social Tags
 Shopping
Recommendation systems generally belong to the  TV Programs, etc.
group of Information Retrieval combined with some Data
Mining and Machine Learning concepts. It is a branch or VI. OBJECTIVES FOR THE IMPLEMENTATION OF
domain under the Data Science field and though progress A RECOMMENDATION ENGINE
has been achieved in this field, still it is growing area of
research. These recommender systems are now of vital There can be several reasons for which businesses
importance in the e-commerce industry. Recommender today use recommendation system: -
system recommend items or products to people such as  To results in an increase in the quantity of products used
books, games, movie/video, electronic product, online by user: In cases of online shopping, this could be
shopping products and other classes of products. executed by “increase the quantity of item or product
Recommender system help the customers to get selling”, whereas in the case of news portal it could be
recommendations that are personalized, additionally translated to “increase in the number/quantity of news
helping users to make apt decisions in the online purchases, items being read by the user”.
rise in the sales and redefining the users browsing  Allow the customer to select item or product that could
experience, retaining the customer base and enhancing their be otherwise difficult to search or locate without an
shopping experience. Often, the “information overload accurate and precise recommendation system: This
problem” is being solved by these search engines, but they functionality allows less-known products and the
do not aim to provide data personalization. customers generally do not tend to use to distinguish
Recommendation engine thus help a step ahead by themselves from the way the user uses, so she can notice
providing personalization. them.
 Increasing the user satisfaction or customer satisfaction
A recommender system is thus a major technique that levels: A well-developed and built recommendation
handles the “information overload problem” of Information system lightens the knowledge of the customer with the
Retrieval by advising its customers with apt and relevant website. Users will generally find the recommendation
items. Today, many recommendation systems are being as interesting, relevant, accurate, precise and with a
built for the domains differently. But the ones developed correctly designed human-computer interaction. It will
are not accurate to satisfy the information needs of the be one which the user can use easily and will also enjoy
customers. Hence, it is obviously necessary to develop or using with the passage of time. This combination
build good, effective and high quality recommendation provided with precise predictions and rich user interface
system. In designing such recommendation systems, will increase the subjective evaluation of these systems.
designers often face issues and variety of challenges that These in return will then increase the system usage and
needs to be addressed with proper attention. the probability that the recommendations will be
accepted. This will also tend to protect the user loyalty.
IV. MOTIVATION  Help in understanding the user taste in a better manner:
From the users’ choices and preferences which has been
In order to analyze and understand the behavior and collected, service providers, developers, business
the tastes of the users, many organizations currently include managers, etc. could need to develop new marketing
recommendation system or recommendation engine in their policy that would suit individual tastes.
websites in order to reach some business goals. Among
these goals we have: VII. IMPLEMENTATION FRAMEWORK
 the rise of traffic to the web sites,
 effective marketing policy elaboration tailored for The proposed system is being built using Python
meeting the customer tastes, language and Django framework.
 or in general the promotion or advertisement of product.
Python is an interpreted, object oriented and high
V. APPLICATION FIELDS FOR level programming language which provides support for
RECOMMENDATION SYSTEMS different module and package. This encourages the
concepts of program modularity and code reuse.
Some of the most widely used areas of
recommendation systems include: - Django is a free, open-source and a widely-used
 Books Python web application framework. Generally, frameworks
 Documents are used to save developers from having to reinvent the
 News wheel and to help reduce some of the need and overhead to
 Research Articles start developing a new website right from the scratch. This
 Images common functionality which are used for building the web
 Movies applications comes with the framework instead of using
 Search Queries separate libraries.

IJISRT19FB208 www.ijisrt.com 445

Volume 4, Issue 2, February – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 Django MVT C. Simple Recommender System
MVT stands for Model-View-Template. It is a The Simple Recommender offers generic
software design pattern being used in Django. It has a recommendations to users based on the popularity of the
collection of three vital components – Models, Views and movie and also based on the genre. The idea involved in
Templates. this is - movies which are popular will have a higher
 Model – It helps to take care of the database. It is a data probability of likes by audience.
access layer that is used for data handling.
 Template – It is presentation layer which handles the All we have to do is sort our movies based on ratings
User Interface (UI) fully. and popularity and display the top movies of our list. As an
 View – It is used to write the logic for main business. added step, we can pass in a genre argument to get the top
The view also interacts with the model to carry the movies of a particular genre.
required data and then renders it to the template.
We can use IMDB's weighted rating formula to
Figure 2 shows the MVT-based flow as - User first construct my chart. It can be represented mathematically as
sends a request for a resource to application. Now, Django follows:
is working as a controller. So, it checks if the resource is
available in the URL. If the URL can map the resource Weighted Rating = ( (v/(v+m)) * R) + ( (m/(v+m)) * C)
correctly, then view is being called that interacts with the
model and the template. This template is then rendered to where,
the user. Django responds to the user and send this template v – total number of votes accumulated for the movie
as a response. m – minimum number of votes required for the movie to
be listed in the charts
R – average rating of the movie
C – mean vote for the report

D. Approach
In the process of building a recommendation engine,
we have implemented Simple Recommender-based
approach with the intention of analyzing the dataset under
consideration.

For executing the results using Python, first we need

to import the required library files and load the dataset into
a pandas data frame. Then we can retrieve the genre for all
data records and if vote count exists. If it exists, we can
retrieve the required values. Similarly, we can check for
vote average. Then we can compute the mean of the above
vote averages obtained and calculate the min number of
Fig 2:- MVT-based control flow in Django votes that is necessary for the movie to be in the chart. Then
we can filter the required movies that are qualifying for the
VIII. RESEARCH METHODOLOGY charts and return the data. A function needs to be defined to
calculate the weighted rating for each movie by defining a
A. Research Goals feature known as weighted rating and calculating it. Thus,
The goal of the Movie Recommendation System is to we can then sort the data set based on this weighted rating.
recommend movies to users which they have not watched
yet. In achieving this goal, following two strategic Additionally, we can specify a genre and the data will
objectives can be addressed: - be displayed based on that particular genre only to the user.

 Objective 1: To identify areas where relevant and IX. RESULTS

accurate predictions could make the most difference.
 Objective 2: To assist users by recommending movies. Simple Recommender based approach: This system
used overall TMDB Vote Count and Vote Averages to
B. Source of Data Collection build Top Movies Charts, in general and for a specific
For this study secondary data has been collected. the genre. The IMDB Weighted Rating System was used to
data relating to The Movies Dataset for the movies released calculate ratings on which the sorting was finally
on or before July 2017. These files contain metadata i.e. performed.
detailed data about the main data for all the thousands of
movies listed in the MovieLens Dataset. Thus if I say, print(pass results.head(15)), the top 15
movies are displayed to the user as shown in Figure 3.

IJISRT19FB208 www.ijisrt.com 446

Volume 4, Issue 2, February – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

I am very thankful to the IT Department and Dr.

Santosh Kumar Sing, Head of Department of Information
Technology for his kind co-operation in the completion of
this project.

Special thanks to my project guides, Dr. (Mr.) Santosh

Kumar Singh and Prof. Mahendra Sharma who have given
me the opportunity to do this, which even helped in doing
Research and I learnt many new things.

Fig 3:- Results displaying the top 15 movies REFERENCES

Thus, the top movie according to the charts is [1] Sidnooma Christian Kabore. S 2012. Design and
“Inception” which has the highest weighted rating as implementation of a recommender system as a
calculated by the system – 7.917588 and “Se7en” which has module for Liferay portal.
the weighted rating of 7.811669 appears to be the one with [2] N. Vaidya and A. R. Khachane. Recommender
the least weighted rating and is basically at the last in the systems-the need of the e-commerce ERA. 2017
charts. International Conference on . Computing
Methodologies and Communication (ICCMC),
Now, if I say, Erode, pp. 100-104. doi:
print(chart_genre(‘Romance’).head(15)), the top 15 movies 10.1109/ICCMC.2017.8282616
are displayed to the user as shown in Figure 4. [3] Khusro, Shah & Ali, Zafar & Ullah, Irfan. (2016).
Recommender Systems: Issues, Challenges, and
Research Opportunities. 10.1007/978-981-10-0557-
2_112.
[4] Venkatesan M. (Government Arts College
(Autonomous), India) and Thangadurai K.
(Government Arts College (Autonomous), India).
History and Overview of the Recommender Systems.
2017. DOI: 10.4018/978-1-5225-0489-4.ch004
[5] Francesco Ricci, Lior Rokach, and Bracha Shapira.
Recommender Systems: Introduction and Challenges.
Springer Science+Business Media New York 2015.
DOI 10.1007/978-1-4899-7637-6_1
Fig 4:- Results when value is passed for a specific genre [6] https://www.analyticsvidhya.com/blog/2015/10/reco
mmendation-engines/
Thus, the top romance movie according to the charts [7] https://en.wikipedia.org/wiki/Recommender_system
is “Dilwale Dulhania Le Jayenge” which has the highest [8] https://www.javatpoint.com/django-mvt
weighted rating as calculated by the system – 8.565285 and [9] www.kaggle.com
“Silver Linings Playbook” which has the weighted rating of [10] www.datacamp.com
6.970581 appears to be the one with the least weighted
rating and is basically at the last in the charts.

X. CONCLUSION

This paper presents a movie recommendation system

using Simple Recommender-based approach using Python.
It inputs the data source for the movies Dataset for movies
data released before July 2017 and provides
recommendations for movies.

ACKNOWLEDGMENT

I would like to express my gratitude to my college,

Thakur College of Science and Commerce (TCSC) for
giving me this opportunity to accomplish the project work.

I express my heartfelt gratitude to the Principal Dr.

(Mrs.) C.T. Chakraborty for her kind, inspiring and
illustrious guidance and ample encouragement.

IJISRT19FB208 www.ijisrt.com 447