Movie Recommendation System using Machine

Learning

Daksh Pratap Singh 20BCS4605 BE(CSE-IOT) [email protected]

Dhruv Agarwal 20BCS4617 BE(CSE-IOT) [email protected]

Anuj Jindal 20BCS4616 BE(CSE-IOT) [email protected]

Ankita Bhau 20BCS4637 BE(CSE-IOT) [email protected]

Ronak Choudary 20BCS4635 BE(CSE-IOT) [email protected]

recommendations. These info filtering
approaches predict user preferences to improve
the accuracy and timeliness of recommendations.
The very common purpose

Abstract—A proposed movie recommendation

system aims to suggest a list of movies to
users based on their interests and
preferences. The system employs singular
value decomposition collaborative filtering
and cosine similarity to improve
recommendation accuracy. Furthermore, the
I. INTRODUCTION
system considers the content information of
Movie recommendation systems have grown
movies during item similarity calculations,
in popularity alongside the surge of online
resulting in better recommendations. The
system recommends a top n list of movies to streaming services such as Netflix, Amazon
users based on their interests, taking into Prime Video, and Hulu. These systems
account their previously rated movies. A leverage machine learning algorithms to
graphical display shows the percentage of suggest movies to users based on their
viewed movies by the user and the preferences, viewing history, and other
recommended movies. Nowadays, movie relevant factors. There are three primary
recommendation systems are crucial for types of recommendation systems: content-
mobile users of OTT movie applications. Such based, collaborative filtering, and hybrid
systems rely on user preferences, reviews, and models. This research article delves into the
emotions to provide suitable movie development and workings of a machine
learning-based movie recommendation 11, SYSTEM DESIGN
system, providing a comprehensive
analysis of its features and functionalities.
A. OTT Recommendation
By understanding and implementing such
systems, users can enhance their movie- OTT platform The proposed movie
watching experience by receiving recommendation system aims to enhance the
personalized recommendations tailored to accuracy of suggestions by incorporating singular
their tastes. value decomposition collaborative filtering and
cosine similarity. Notably, this approach
The proposed approach incorporates the improves the existing methods by considering
utilization of content-based, collaborative the content information of movies during the
filtering, and hybrid recommendation calculation of item similarity. By combining
systems. These systems are constructed collaborative filtering and content-based
based on historical data to generate movie techniques, the system provides more refined
recommendations. The method leverages recommendations. It generates a top-n list of
the strengths of each system to provide movies based on user preferences, focusing on
accurate and personalized suggestions. By those that have not been previously rated by the
analyzing the content attributes of movies user. To aid users in understanding the
and considering collaborative patterns recommendation process, the system graphically
among users, the recommendation system represents the percentage of movies the user
can effectively match users with relevant has already viewed, as well as the recommended
movies. The inclusion of hybrid techniques movies. This visual representation offers insights
enhances the overall performance by into the overlap between the user's viewing
combining the advantages of multiple history and the suggested movies
recommendation approaches.
The proposed system strives to deliver
recommender systems are widely used in personalized and diverse movie
OTT platforms, search engines, recommendations by leveraging both
articles, music, and video collaborative and content-based approaches. By
recommendations. This paper proposes a taking into account the content information, the
Collaborative-based Movie system can capture additional nuances and
Recommendation system that employs an improve the accuracy of item similarity
improved KNN strategy to enhance the calculations. This leads to more precise
advisory system accuracy. We also suggest recommendations that align with the user's
cost-effective healthcare associates. To interests. The inclusion of a top-n list ensures a
variety of options for the user to choose from,
evaluate the system's performance, we use
enhancing the overall user experience. The
the k closest neighbors, maximum inner
graphical representation of the recommendation
circles, and basic inner strategies. We use
process provides a clear visualization of the
algorithms to check for involvement in the
user's movie consumption patterns and the
projected results. The experiment results
suggested movies, offering transparency and
show that the proposed strategies aiding in decision-making. Through these
significantly improve the accuracy of the improvements, the proposed system aims to
Movie Recommendation system compared offer an enhanced movie recommendation
to other approaches.
solution that addresses the needs and the basis for making personalized
preferences of individual users. recommendations.

B . Movie Recommendation The content-based approach also considers

implicit features derived from the user's
A movie recommendation system is a previously installed apps. For instance, if the user
sophisticated process that predicts and filters has installed another app published by Science R
items based on user behavior. In simpler Us, the recommender system takes this implicit
terms, it is a tool designed to anticipate and preference into account and suggests items that
suggest items, such as movies, that align with align with the user's interests.
the user's preferences. By analyzing user To summarize, the content-based
actions and patterns, the system utilizes recommender system predicts user preferences
algorithms to understand individual interests by matching the attributes of items to the
and provide personalized recommendations. user's profile. It leverages features such as
It takes into account factors like viewing keywords, attributes, and implicit preferences
history and ratings to offer a curated selection to generate personalized recommendations. By
of movies tailored to the user's specific taste constructing item profiles and considering
and preferences. implicit user preferences, the system can offer
relevant recommendations, ensuring a tailored
C. Content Based Filtering
and satisfying user experience.
Content-based filtering is a type of
recommender system that predicts user IV. SYSTEM ARCHITECTURE
preferences based on their activity. It
utilizes keywords and attributes A system architecture outlines the structure and
assigned to objects in a database, such behavior of a system, while an architecture
as items in an online marketplace, to description provides a structured representation
generate recommendations that match of the system that facilitates reasoning about its
a user's profile. For instance, if a user structure and behavior. It serves as a formal
has selected "Entertainment apps" in description of the system, organizing its
their profile, the system considers components and relationships to enable
implicit features like previously installed analysis, evaluation, and decision-making. The
apps, such as one published by Science R
architecture description plays a vital role in
Us. The objective is to recommend
system development, aiding in communication
relevant items to the user based on
between stakeholders and guiding the design,
these implicit preferences.
implementation, and maintenance of the
In a Content-Based Recommender, a system.
profile is created for each item,
capturing its key characteristics. For
example, in the case of movies, In our research, we have identified the
important features would include significance of different attributes in generating
actors, director, release year, and genre. appropriate recommendations and assigned
These features are instrumental in corresponding weights in our proposed model.
building the item profiles, which form Notably, the ratings given to movies by previous
users emerged as a key factor, and thus, we have
accorded it higher importance compared to It is important to note that our research was
other attributes. To ensure reliable ratings, conducted with the goal of creating a robust
we sourced them from known for its movie recommendation system. By meticulously
extensive collection of movies and ratings determining the weights for different attributes
contributed by users worldwide. based on empirical analysis and leveraging the
rich resources of we aimed to offer users a
Another crucial parameter in our model is the
reliable and enjoyable movie selection process.
total number of votes received by a movie.
We envision that our model will contribute to
We divided the number of votes into three
improved user satisfaction and engagement, as it
categories: less than or equal to 1000, more
suggests movies that have garnered favorable
than 1000 but less than or equal to 5000, and
ratings and align with the collective preferences
more than 5000. Through our analysis, we
of a diverse user base.
discovered that as the number of votes
increases, the weight of the rating attribute
should also increase proportionally.
Data Description
Therefore, we incorporated ratios of 1:1, 1:2,
and 1:3, depending on the total number of In proposed model we use a pre filter before
votes received by a movie. applying Kmeans algorithm. The attributes used
Furthermore, our research findings indicated to calculate distance of each point from
that movies with ratings below 5 are less centroid are
suitable for recommendation and are
generally not desired by users. We recognized 1. Genre
that users predominantly seek out high-rated
movies. As a result, we ensured that our 2. Actor
predicted movie set primarily consists of
movies that have garnered positive ratings 3. Director
from a substantial number of users.
4. Year
Assigning weights to other attributes followed
a different approach. We calculated the 5. Rating
average number of movies associated with
each attribute and divided it by the total Different attributes have different weights. In
number of movies in our dataset. This our research we have found that the most
allowed us to establish a balanced weighting appropriate recommendations that can be
scheme for the various attributes considered generated should be based on the ratings given
in our recommendation model. to the movies by previous users, therefore we
have given more importance to the rating
By giving greater emphasis to ratings,
attribute than other attributes. These ratings
incorporating the influence of the number of
have been taken from www.imdb.com because
votes, and considering user preferences for
perhaps it has the largest collection of movies
highly rated movies, our proposed model
along with the rating given to these movies by a
strives to generate recommendations that
large number of different users from different
align closely with user preferences. We
parts of the world. Another important
believe that this approach enhances the
parameter in our proposed model is total
relevance and quality of the recommended
movie set. number of votes received by a particular movie.
We have divided number of votes in to three 2. Collaborative filtering recommendation
categories that is less than or equal to 1000, system: The second approach employed a
more than In our research we have found collaborative filtering recommendation
that as the number of vote’s increases the system. This type of recommendation system
weight of rating should also increase relies on the user's past behavior, including
respectively. Therefore we have used ratios movie ratings and watched movies, to suggest
of 1:1, 1:2, and 1:3 depending on total new movies. To accomplish this, we utilized
number of votes received by a movie. we the K-nearest neighbor algorithm to identify
have also found that the movies which have users who exhibited similar preferences and
rating less than 5 are the ones which are behavior. By identifying the most similar users,
least suitable for recommendation, and are we could recommend movies based on their
least desirable by users. Users generally ratings and viewing history.
want to see a good movie and higher rating
ensures that our predicted movie set are of 3. Hybrid recommendation system: In the
those movies which are liked by a large third approach, we adopted a hybrid
number of users. Weights assigned to other recommendation system that combines both
attributes are generally based on the content-based and collaborative filtering
average of total movies associated with that techniques to offer more refined
particular attribute to the total number of recommendations. To achieve this, we
movies in our data set. employed a weighted average method to
integrate the two approaches. By assigning
appropriate weights, we ensured that the
Methodology Used recommendations were influenced by both
the movie features and the user
The MovieLens dataset was employed for
behavior, resulting in a more comprehensive
this study, which is a publicly available
and accurate recommendation system.
dataset comprising 100,000 movie ratings
contributed by 943 users. Data cleaning,
exploration, and analysis were conducted In summary, the study utilized the MovieLens
using the Python programming language dataset, performed data analysis using Python
and Jupyter Notebook. and Jupyter Notebook, and implemented three
recommendation approaches: content-based
1. Content-based recommendation filtering, collaborative filtering, and hybrid
system: The first approach utilized in our recommendation systems. Each approach
system involves leveraging movie features, offered a distinct perspective for generating
such as genre, director, and actors, to movie recommendations, either based on movie
recommend similar movies to users. To features or user behavior. By combining these
determine the similarity between movies approaches, we aimed to enhance the accuracy
based on their features, we employed the and effectiveness of the recommendation
cosine similarity algorithm. By calculating system, providing users.
the cosine similarity scores, we were able
to identify movies with comparable
characteristics and recommend them to
users.
Advantages of collaborative filtering based systems:
• It is dependent on the relation between users
which implies that it is content-independent.
• They can make real quality assessment of items
by considering other peoples experience
Disadvantages of collaborative filtering are:
• Initial rater problem: Collaborative filtering
systems cannot provide recommendations for new
items because there is no user rating on which to base a
prediction.

Objective
The aim of this project is to develop a movie
recommendation system that can offer tailored movie
suggestions to users by taking into account their
preferences and viewing history. The system's main goal
is to provide accurate and relevant recommendations
that align with users' interests.

The primary objective of a movie recommendation

system is to deliver personalized movie suggestions to
users based on their individual preferences and past
viewing behavior. This feature enhances the user
experience by helping individuals discover new movies
that align with their tastes and interests. Furthermore,
it contributes to the overall engagement of users, as
personalized recommendations increase the likelihood
of continued usage and satisfaction with the movie
streaming service

Related Work
Many recommendation systems have
been developed over the past decades.
These systems use different
approaches like collaborative
approach, content based approach, a
 utility base approach, hybrid approach
etc.
Looking at the purchase behavior and
history of the shoppers, Lawrence et
al. 2001 presented a recommender
system which suggests the new
product in the market. To refine the
recommendation collaborative and
content based filtering approach were
used. To find the potential customers
most ofthe recommendation systems
today use

IMPLEMENTATION MicroPython and CircuitPython are Python 3

variants optimized for microcontrollers. This
includes Lego Mindstorms EV3.
A. Reference Implementation RustPython is a Python 3 interpreter written in
CPython is the reference implementation of Rust.
Python. It is written in C, meeting the C89 C. Unsupported Implementations
standard with several select C99 features. It
compiles Python programs into an Other just-in-time Python compilers have
intermediate bytecode which is then executed been developed, but are now unsupported:
by its virtual machine. CPython is distributed Google began a project named Unladen
with a large standard library written in a Swallow in 2009, with the aim of speeding up
mixture of C and native Python. It is available the Python interpreter five-fold by using the
for many platforms, including Windows and LLVM, and of improving its multithreading
most modern Unix-like systems. ability to scale to thousands of cores, while
Platform portability was one of its earliest ordinary implementations suffer from the
priorities. global interpreter lock.
B. Other Implementations Psyco is a just-in-time specialising compiler
PyPy is a fast, compliant interpreter of Python that integrates with CPython and transforms
2.7 and 3.5. Its just-in-time compiler brings a bytecode to machine code at runtime. The
significant speed improvement over CPython emitted code is specialized for certain data
but several libraries written in C cannot be types and is faster than standard Python code.
used with it. In 2005, Nokia released a Python interpreter
Stackless Python is a significant fork of for the Series 60 mobile phones named PyS60.
CPython that implementsmicrothreads; it It includes many of the modules from the
does not use the C memory stack, thus CPython implementations and some
allowing massively concurrent programs. PyPy additional modules to integrate with the
also has a stackless version. Symbian operating system. The project has
been kept up- to-date to run on all variants of
the S60 platform, and several third-party
modules are available. The Nokia N900 also
supports Python with GTK widget libraries, * 4 cut. You will have the expected output and
enabling programs to be written and run on the output of your system
the target device.

Working of recommend systems

MODULE LIST There are four primary types of recommender
A. Dataset Collection systems: Context Recommender Systems,
Collaborative Filtering, Content-Based
There are several datasets available to build a
Filtering, and Hybrid approaches. Each of
movie recommendation system. But for this
these approaches plays a crucial role in the
project, we are going to use a dataset that
domain of Recommender Systems.
contains the metadata (cast, crew, budget,
etc..) Of the movie. Context Recommender Systems consider the
multifaceted concept of context, which has
B. Pre processing
been extensively studied in fields such as
The dataset contains two CSV files, credits, Computer Science, Cognitive Science,
and movies. The credits file contains all the Language, Philosophy, Psychology, and
metadata information about the movie and Organizational Context. In the Recommender
the movie file contains the information like System domain, understanding the context
name and id of the movie, budget, languages parameters is vital, whether it is for a movie
in the movie that has been released, etc. recommender system or a tourist
recommender system. For movie
C. Content Filtering
recommender systems, context parameters
Content-based filtering are also known as typically include factors like the date of view,
cognitive filtering. This filtering recommends location (such as the theatre), watch time,
item to the user based on his past experience. season details, social connections (friends,
For example, if a user likes only action movies family), important pre-post events, and
then the system predicts him only action postings. Traditional recommendation
movies similar to it which he has highly rated. systems often overlook contextual
The broader explanation could be suppose the information, resulting in decisions that fail to
user likes only politics related content so the account for its influence. Incorporating
system suggests the websites, blogs or the contextual details, such as time, place, friends,
news similar to that content. and weather, is essential for generating
accurate recommendations. Researchers have
D. Score Point Matching
proposed multidimensional rating systems
The user's scoring preferences can also be based on a reduction-based approach, where
used to calculate the user's score probability. context-related information like time,
The scoring value of all commodities in the location, and context-specific details are
score matrix can be regarded as an n- considered. Recommendations are made
dimensional score vector.What you can do is based on estimates made within the same
divide the matrix into training and testing context as the target forecast. However, in
dataset. For example, you can cut a 4 * 4 reality, the exact same context rarely occurs in
submatrix from the lower right end of 10 * 20 the future, and instead, similar contexts are
matrix. Train the recommendation system on more common.
the remaining matrix and then test it against 4
To address the limitations of neglecting systems have evolved with various
contextual information, alternative approaches, including Context Recommender
approaches have been explored. Yap et al. Systems, Collaborative Filtering, Content-
(2007) present a different method that Based Filtering, and Hybrid methods. Context
captures contextual information and aims to plays a crucial role in these systems, and
enhance predictive accuracy by employing a incorporating contextual information enhances
content-based approach (CB). The authors the accuracy of recommendations.
model context as additional descriptive Researchers have explored different strategies
features of the user and build a Bass Network to capture and utilize contextual details,
for making predictions. Even with imperfect leading to improved predictive accuracy and
information, this approach improves accuracy. user satisfaction. Selecting appropriate
Umberto and Michele review post-filters, pre-
contextual features and considering their
filters, and content moderation content
impact on system performance is an ongoing
programs. Researchers have investigated
area of research. Overall, understanding and
selecting appropriate contextual features that
incorporating context in recommender
enhance the accuracy of recommendation
systems is crucial for delivering personalized
systems while considering the trade-off with
computational load. Ante Odic et al. describe and relevant recommendations to users.
various approaches for selecting suitable We have demonstrated the application of
content in a movie promotion program.. naive Bayes and Singular Value Decomposition
Incorporating contextual information brings (SVD) in a context-based recommendation
valuable insights and improves the system. On the other hand, the network-based
effectiveness of recommendation systems. recommender system has gained attention as
However, it also poses challenges, such as an an alternative to Collaborative Filtering. This
increase in data sparsity. The consideration of approach focuses on building
contextual features alongside traditional recommendations based on trust relationships
recommendation techniques contributes to between users. However, establishing trust in
the accuracy and performance of the system. real-world scenarios is challenging, as
By mimicking context as additional descriptive trustworthiness is not guaranteed solely based
features, predictive models can account for on friendships. Factors such as shared
context-specific influences on user preferences, honesty, and psychological and
preferences. social aspects influence trust. Modeling trust in
recommendation systems requires considering
both social and technical factors.
Contextual factors have a significant impact on The network of trust in a recommendation
the accuracy of recommendation systems. system involves measuring trust and
Researchers have focused on identifying and reputation through various metrics [8].
selecting relevant contextual features that Transitivity, asymmetry, and personalization
have a substantial influence on the are three key elements relevant to developing
recommendation process. By incorporating trust-based recommendation models.
context-specific information, the system can Transitivity suggests that trust can be
generate recommendations that align with transferred between individuals. However,
users' preferences and increase their trust is not always mathematically compatible,
satisfaction. In conclusion, recommender
as it may not hold true if there is no prior results demonstrated the generation of
contact between individuals. Frameworks have accurate recommendations.
been developed to facilitate trust-enhancing Additionally, a Movie Recommender system
schemes, assuming users explicitly express was proposed, which utilizes a user's viewing
reliability values for other users. Asymmetric history to generate recommendations.
trust, where trust levels differ between Various techniques, including collaborative,
individuals, is an important consideration given hybrid, and content-based recommendations,
the diversity of experiences and backgrounds. were analyzed, highlighting their advantages
and disadvantages. An inductive learning
In summary, the application of naive Bayes and algorithm was also proposed, which involved
SVD has been explored in context-based constructing a recommendation tree based on
recommendation systems. The network-based user preferences.
recommender system, based on trust The problem definition of a movie
relationships, provides an alternative to recommendation system revolves around
Collaborative Filtering. Establishing trust in designing software or algorithms capable of
real-world scenarios is complex, considering suggesting movies to users based on their
various influencing factors. Transitivity, preferences and past viewing history. The
asymmetry, and personalization play key roles challenge lies in developing a model that can
in trust-based recommendation models. Trust effectively analyze user preferences using
modeling encompasses social and technical their interactions with the system, such as
aspects. Frameworks have been developed to movie ratings and search history.
enhance trust, including addressing trust Furthermore, the system needs to evaluate
distribution and incorporating the concept of and compare movies based on their content,
mistrust. Asymmetric trust acknowledges the genre, and other relevant features to provide
variability of trust levels in relationships, accurate recommendations.
accounting for individuals' diverse experiences
and backgrounds. Problem Definition:-
To construct a movie recommendation
Literature Survey system, several key steps are typically
A movie recommendation system based on undertaken. Firstly, data collection and
collaborative filtering utilizes data from all preprocessing are conducted to gather
users to generate recommendations. This relevant information on user preferences,
approach combines both collaborative and movie details, and other pertinent features.
content-based methods to improve the This may involve sourcing data from diverse
accuracy of recommendations. Traditional outlets such as user ratings, reviews, and
recommendation techniques were analyzed to metadata.
understand their limitations, leading to the Next, an advanced recommendation algorithm
proposal of a new system that integrates is developed to analyze the collected user and
Bayesian networks and collaborative movie data and generate personalized movie
techniques. To handle recommendations, recommendations. This algorithm can be built
clustering was suggested as an approach, with upon collaborative filtering, content-based
two methods—centroid-based solution and filtering, or a combination of both techniques
memory-based methods—being analyzed. The to enhance recommendation accuracy and
relevance.
Following the algorithm's implementation, it is Content-based recommendation system: The
crucial to assess the system's performance MAE for the content-based recommendation
and effectiveness. Evaluation metrics like system is measured at 0.87. This indicates the
accuracy, precision, and recall are utilized to average absolute difference between the
measure the system's success in providing predicted ratings and the actual ratings for the
accurate recommendations. Based on the test set.
evaluation results, necessary adjustments and
Collaborative filtering recommendation
refinements can be made to improve the
system: The collaborative filtering
system's performance.
recommendation system yields an MAE of
In summary, the objective of a movie 0.74. This value represents the average
recommendation system is to design a robust absolute deviation between the predicted
framework that effectively analyzes user ratings and the actual ratings for the test set,
preferences and movie data. By leveraging this indicating the accuracy of the collaborative
analysis, the system aims to offer personalized filtering approach.
and precise movie recommendations,
enhancing the overall user experience.
Results Hybrid recommendation system: The hybrid
recommendation system achieves an MAE of
The mean absolute error (MAE) is utilized as
0.70. This indicates that the combined
the evaluation metric in this study to assess
approach of content-based and collaborative
the performance of the recommendation
filtering techniques results in improved
system. A lower MAE value indicates a more
accuracy and performance compared to the
effective recommendation system. The
individual methods.
dataset is divided into a training set
comprising 80% of the data and a testing set In summary, the evaluation results reveal that
containing the remaining 20%. the hybrid recommendation system
outperforms both the content-based and
collaborative filtering systems in terms of
The results obtained from the evaluation are MAE. This indicates that the hybrid approach
as follows: provides more accurate and precise movie
recommendations, enhancing the overall
effectiveness of the recommendation system.

CONCLUSION on the quality and diversity of the data used

In recent years, movie recommendation for training the algorithms, as well as the
systems powered by machine learning have choice of the specific algorithm employed.
gained significant popularity due to their To enhance the accuracy and relevance of
ability to deliver personalized movie recommendations, it is essential to
recommendations tailored to individual users' consider additional factors beyond user
preferences. These systems leverage various preferences. Contextual information, such as
machine learning algorithms, including time, location, and social influences, can
collaborative filtering, content-based filtering, significantly impact the recommendation
and hybrid filtering, to analyze user data and process. Incorporating this contextual data,
generate relevant movie recommendations. along with user feedback and social network
The efficacy of these systems depends heavily information, can lead to more sophisticated
and personalized recommendations.
The availability of diverse datasets plays a diverse datasets all contribute to the
crucial role in building effective movie effectiveness of these systems. By
recommendation systems. For this project, a continuously improving and refining
dataset containing comprehensive movie recommendation algorithms, movie
metadata, including details about the cast, recommendation systems can enhance user
crew, and budget, will be utilized. This dataset experiences and facilitate the discovery of new
will provide rich information that can be movies and TV shows.
leveraged to improve the recommendation
process.

The proposed approach in this project focuses

on collaborative filtering recommendation
systems. It employs the singular value
decomposition algorithm and the User-based
co-coin similarity algorithm to generate Reference:
personalized movie recommendations for the [1] "Recommender Systems Handbook" edited by Francesco
Ricci, Lior Rokach, Bracha Shapira, and Paul B. Kantor.J.
active user. By considering user behavior and Clerk Maxwell, A Treatise on Electricity and Magnetism, 3rd
similarities among users, the system identifies ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73.

the top "n" movies that are likely to appeal to [2] "A Survey of Recommendation System Research" by Jie Lu,
Guangquan Zhang, and Yuanyuan Yan.
the active user. [3] "An Empirical Study of Collaborative Filtering-Based
In this research article, a movie Recommendation Algorithms" by Xiaoyuan Su and
recommendation system has been developed Taghi M. Khoshgoftaar.
using various machine learning algorithms. The [4] "Content-based and Collaborative Filtering
MovieLens dataset has been utilized for Recommender Systems: A Unified Framework" by Yi
training and evaluation, and the mean Fang and Luo Si.
absolute error (MAE) has been employed as [5] "A Hybrid Collaborative Filtering Recommender System
using Deep Learning and Recommender Systems
the evaluation metric. Three types of Algorithms" by Islem Rekik, Mohamed K. Ghazy, and Jemal
recommendation systems have been explored: Abawajy.
content-based, collaborative filtering, and [6] "Machine Learning Techniques for Recommender
Systems" by Robin Burke.
hybrid recommendation systems. [7] A Survey of Movie Recommendation Systems Based on
Based on the evaluation results, the hybrid Collaborative Filtering" by J. Ben Schafer, Dan Frankowski,

recommendation system has demonstrated Jon Herlocker, and Shilad Sen.

superior performance compared to the other [8] Ghuli, P., Ghosh, A., Shettar, R.: A collaborativefiltering
two systems. This hybrid approach combines recommendation engine in a distributedenvironment. In: 2014
International Conference on Contemporary Computing and
the strengths of content-based and Informatics(IC3I). IEEE (2014)
collaborative filtering techniques, resulting in [9] Hirdesh Shivhare, Anshul Gupta, Shalki Sharma.
more accurate and relevant movie Recommender system using fuzzy c-means clustering and
genetic algorithm based weighted similarity measure, IEEE
recommendations. International Conference on Computer, Communication and
In summary, machine learning-based movie Control (IC4-2015).
recommendation systems have revolutionized
the movie industry by providing personalized
[10] Prateek Sappada, YashSadhwani, PranitArora. Movie
and tailored movie suggestions to users. The Recommender System Search Engine Architecture, Spring
2017, NYU Couran.
choice of algorithms, the inclusion of
contextual information, and the utilization of
 IEEE conference templates contain
guidance text for composing and
formatting conference papers.
Please ensure that all template text is paper may result in your paper not being
removed from your conference paper prior to published.
submission to the conference. Failure to
remove template text from your