Movie Recommendations System Using ML

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Movie Recommendations System Using Ml

1 3
Shivam Sharma, Shikhar Gaur,
Department of electronics & communication Engineering, Department of electronics & communication Engineering,
KIET Group of Institution’s Ghaziabad, India KIET Group of Institutions Ghaziabad, India
2 4
Shubham Shukla, Vibhav kumar Sachan,
Department of electronics & communication Engineering, Department of electronics & communication Engineering,
KIET Group of Institutions Ghaziabad, India KIET Group of Institutions Ghaziabad, India
Abstract:- A recommendation engine is a type of data providesthe result and various other options like the user's
filtering technology that uses machine learning interest.
techniques to provide the most relevant
recommendations to a particular user or client. It For ex:- if a user wishes to watch a superhero movie,
operates by searching for patterns in client behavior then the recommendation system will recommend all top
data, which may be collected explicitly or implicitly. It movies related to superheroes.
first records a client's area of interest and then uses that
information to enable product recommendations for II. RELATED WORK
customers who appear to be shopping. For example: An
e-commerce website won't know anything about a visitor The research of recommender systems has expanded
if they are completely new to it. So how would the greatly during the past few years. Several methodologies are
positioning strategy advocate the product to the used by recommendation systems to provide useful
consumer in this situation? One practicalsolution may be recommendations. Both collaborative filtering and content-
to suggest the item that is in great demand, or the one based filtering are typically utilized. A user is given a
that is popular. Another practical option is to recommendation for an item via a content- based filtering
recommend the product that will likely bring more system based on their own preferences.
profit to the company. Recommendation engines can be
implemented by using 3 strategies: - Collaborative There are several movie recommendation algorithms
filtering (focuses on collecting and analyzing data about available nowadays. To provide the most accurate
user behavior, preferences, and activities to predict what suggestions, the developers of these systems collect data
a person would like based on how they are similar to from social media to determine consumerpreferences.
other users.), content-based filtering (which works on the
principle that if you like one thing, you'll like this other In this work, we provide a recommendation system
thing, too) and hybrid models. that evaluates the user's preferences to suggest the best
option. The quality and content are considered via
Keywords: Recommendation System, Collaborative Filtering collaborative filtering and content-based filtering. Our goal
Approach, And Content- Based Filtering Method. is to deliver suggestions to users by taking the search
experience into careful, organized consideration.
I. INTRODUCTION
 Technique Used
Recommendation algorithms are present everywhere Our movie recommendation engine employs content-
these days. Several of the biggest IT companies including based filtering as its method. The ratings and reviews that
Netflix, Amazon, Flipkart, Google engage with everyday users submit are used in a content-based movie
use recommendation systems. recommendation system. A user profile is constructed using
this data, and it is then used to suggest movies to both the
Recommender systems are commonly used in various existing user and new users.
fields including movies, music, news, books, articles and
various other products. This filtering technique provides more accurate and
better results if the user gives more inputs on the
The main objective of this system is to filter the data recommendation system.
and provide the result to user in terms of a particular item
with specified domain. The domain- specific item for this
particular system is movies, so our recommendation system
mainly focuses on filtering and predicting the movies based
on the interests and data provided by the user. There are
several methods to develop a movie recommendation
engine, but we built ours using a content-based filtering
strategy. With the help of content-based filtering our system
IJISRT23MAY293 www.ijisrt.com 855

ISSN No:-2456-2165
Fig 3 Cosine Distance
Fig 4 Formula for Finding Similarity among the Data
This formula aids in determining whether two vectors

projected on a multi-dimensional planeare identical.
 Proposed System
Fig 1 Content Based Filtering Basic architecture of our workflow in making this
system is:
Content-based method filtering utilizes Cosine angle
similarity and Term frequency to get the data as shown
in figure 1.
 Term Frequency (Tf)

Term Frequency refers to the frequency ofa particular word
in the data. TF is a statistical measure which determines
howrelevant a word is to a document in thedatafile.
The weight or occurrence of a word in a document can be

counted by using the equation below:
Fig 2 Formula for Filtering the Data
 Cosine Angle Similarity

Cosine angle similarity is a numeric value which
ranges between 0 to 1. It is used to determine the similarity
of two different items from 0 to 1 in regard for one another.
Fig 5 Methodology Used
This score is calculated by comparing the texts of the
two documents in the data. Cosine Similarity can be
The brief description on above blocks are as follows:
measured betweentwo texts irrespective of their sizes.
 Dataset
The cosine angle between two vectors projected in
The dataset is the most crucial component of a
multi-dimensional space is determined by cosine similarity.
machine learning process. So, in our movie recommender
system we have used “TMDB 5000 Movie Dataset”. This
dataset contains Hollywood movies released on or before
January 2018.

ISSN No:-2456-2165
This data consists of cast, crew, genre, plot keywords,
budget, revenue, posters, releasedates, popularity.
Fig 9 Syntax of Linking the Dataset with the System Code
 Preprocessing
In machine learning, we use pandas and NumPy
libraries for preprocessing.
Fig 6 Graph of Movies Dataset Based on Popularity
We may also use ratings to represent the dataset; to

create the graphic, we utilize IMDB's weighted rating
system.
The equation is represented as:

Fig 10 Numerical Analysis
Numeric Python is referred to as NumPy. It is a

library for working with arrays in Python.
Fig 7 Formula for Calculating WR
Moreover, it may be used with algebra, matrices, and
the Fourier transform.
 Above Variable Represents:
 v represents vote totals for the film.
Python's Pandas package is used to manipulate
 m represents minimal number of votes needed to get
datasets. Data analysis, cleansing, exploration, and
seen in the chart.
manipulation are performed by this library function.
 R represents mediocre rating of film C represents mean
vote.
Fig 11 Libraries Used
The most practical and reliable machine learning

library is Scikit-learn (Sklearn). This Python library was
created using Numpy, Scipy, and Matplotlib.
Sklearn assists with graph representation of the

datasets..
A collection of text documents is transformed into a

matrix of token counts using the CountVectorizer function
from the Sklearn toolkit.
Fig 8 Calculated Weighted Rating for the Movies in the

Dataset
Fig 12 CountVectorizer Syntax

ISSN No:-2456-2165
CountVectorizer returns Scipy sparse matrix and this As we have passed ‘Batman Begins’ as the input the
matrix is then converted into a Numpy array. recommendation model has given us recommended movies
related to Batman in the form of output.
Stemming or NLTK it is the process of reducing a
word into its base form by removing affixes from them. For  Website
example data [dance, dancing] will be reduced to [dance, In this step the recommendation model is converted
dance] with help of stemming. into a website with the help of python IDE pycharm.
Streamlit library of python is used for the creation of this
For stemming process NLTK library is used which website.
stands for National language toolkit.
Fig 15 Pickle Library Syntax
Pickle library is used to connect recommendation

system model with website.
Fig 13 NLTK Library Syntax

 Model
After performing all the preprocessing steps the
recommendation model is completed.
 Deploy
This website is deployed on the Heroku server. Three
files are created in order to run the Streamlit app on the
Heroku server those three files are ‘procfile’, ‘steup.sh’ and
gitignore.
“pip freeze > requirements.txt” command is used to

generate all the required libraries to run the website on
Heroku server.
Fig 14 Recommendation System Code
Hence, this recommend function returns the top

recommended movies after passing the title of the movie as
an input.

ISSN No:-2456-2165
III. RESULTS
After the implementation and completion of all the steps the model is tested to check if its working properly or not.
Fig 16 Homepage
This is the homepage (landing page) of our website user can type the name of the movie in search box for getting
recommendations.
Fig 17 Search
Here, user is searching for a movie named “Batman”.

ISSN No:-2456-2165
Fig 18 Result (Recommended Movies)
As we can see, user got the recommendations related to REFERENCES

movies named “Batman”.
[1]. Bilge, A., Kaleli, C., Yakut, I., Gunes, and Polat: A
IV. CONCLUSION study on collaborative filtering techniques.
International Journal of Software Eng. Knowl. Eng.
This study describes a content-based filtering- based 23(08), 1085–1086 (2013) CrossRef with Google
approach for making movie recommendation system. Scholar.
Making the recommendation system effective was the main [2]. Regarding the identification of fake binary ratings:
goal. The main objective is to create a system that can offer Okkalioglu, M., Koc, M., Polat, H. SAC 2015, pp.
users high-quality suggestions without needing a lot of 901–907 in Proceedings of the 30th Annual ACM
personal information, search history, etc. The results of the Symposium on Applied Computing. America, ACM
trials show that the recommended technique provides (2015).
consumers with relevant suggestions. [3]. Kaleli, C., and Polat, H.: Privacy-preserving naive
bayesian classifier based upon on distributed data.
 Future Enhancements Computer Intelligence 31(1):47-68 (2015)
Future work will involve using user ratings of movies [4]. Hongli Lin, Xuedong Yang, and Weisheng Wang. A
on websites like Rotten tomatoes, IMDb, etc. We will try to content-boosted collaborative filtering algorithm for
use Hybrid filtering technique in our model to get better personalized training in interpretation of radiological
predictions and recommendations. imaging. Journal of digital imaging. 107-13 (2014)
[5]. Harpreet Kaur Virk, Er Maninder Singh, and A
Furthermore, we will try to add the data of the movies Singh. Analysis and design of hybrid online movie
from various film industries. recommender system. International Journal of
Innovations in Engineering and Technology Volume,
We will also work on improving and updating the user 5,2015.
interface of our website in order to make our website user [6]. Urszula Ku zelewska. Recommendation system
friendly. engines. Iranian Journal of Energy and
Environment, 2019.
ACKNOWLEDGEMENT
The authors would like to express gratitude towards

KIET Group of Institutions and Dr. Shubham Shukla sir
(ECE Dept.) fortheirfinancial and technical guidance.

Movie Recommendations System Using ML

Uploaded by

Copyright:

Available Formats

Movie Recommendations System Using ML

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Movie Recommendations System Using ML

Uploaded by

Copyright:

Available Formats

Volume 8, Issue 5, May – 2023 International Journal of Innovative Science and Research Technology

Movie Recommendations System Using Ml

IJISRT23MAY293 www.ijisrt.com 855

Fig 3 Cosine Distance

Fig 4 Formula for Finding Similarity among the Data

This formula aids in determining whether two vectors

 Term Frequency (Tf)

The weight or occurrence of a word in a document can be

Fig 2 Formula for Filtering the Data

 Cosine Angle Similarity

IJISRT23MAY293 www.ijisrt.com 856

Fig 9 Syntax of Linking the Dataset with the System Code

Fig 6 Graph of Movies Dataset Based on Popularity

We may also use ratings to represent the dataset; to

The equation is represented as:

Numeric Python is referred to as NumPy. It is a

Fig 11 Libraries Used

The most practical and reliable machine learning

Sklearn assists with graph representation of the

A collection of text documents is transformed into a

Fig 8 Calculated Weighted Rating for the Movies in the

Fig 12 CountVectorizer Syntax

IJISRT23MAY293 www.ijisrt.com 857

Fig 15 Pickle Library Syntax

Pickle library is used to connect recommendation

Fig 13 NLTK Library Syntax

“pip freeze > requirements.txt” command is used to

Hence, this recommend function returns the top

IJISRT23MAY293 www.ijisrt.com 858

Here, user is searching for a movie named “Batman”.

IJISRT23MAY293 www.ijisrt.com 859

Fig 18 Result (Recommended Movies)

As we can see, user got the recommendations related to REFERENCES

The authors would like to express gratitude towards

IJISRT23MAY293 www.ijisrt.com 860

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.