Proceedings of the Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV 2021).

IEEE Xplore Part Number: CFP21ONG-ART; 978-0-7381-1183-4

A Review on Movie Recommendation System Using

2021 Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV) | 978-1-6654-1960-4/20/$31.00 ©2021 IEEE | DOI: 10.1109/ICICV50876.2021.9388619

Machine Learning
Smitha N1,Anusha D2,Chaithanya C3,Sindhu J4,Tanuja R5,Hemanth Kumar H S6
1
Assistant Professor, Department of Computer Science & Engineering, CMR Institute of Technology, Bengaluru, India
2,3,4
Department of Computer Science & Engineering, CMR Institute of Technology, Bengaluru, India
5
Department of Computer Science &Engineering, University Visvesvaraya College of Engineering, Bangalore University,
Bengaluru, India
6
Assistant Professor, Department of Computer Science & Engineering, CMR Institute of Technology, Bengaluru, India

Email: 1smithan.ckm@gmail.com; 2anus17cs@cmrit.ac.in;3 chai17cs@cmrit.ac.in ;4 sind17cs@cmrit.ac.in;

5 r_tanuja@yahoo.com;6hemanth.kit@gmail.com

Abstract—System which gives recommendations usually 1.Content based filtering: It thoroughly relies upon the
filters the given data using various methodology and suggests the similarity of an item, then these products are highly
relevant one to the customer benefit. In day-to-day life people recommended and considered [5].
usually use a powered recommendation system in many areas
like movies, books, music, news, items etc. In this paper a wide 2.Memory based Collaborative filtering: It is relying on the
range of work is reviewed in the field of a recommender system relation between users and its items and it totally depends on
for movies where dataset source, methods used and accuracy are behavior but not the context.
compared to deduce best one and future scope for improvement
in this area are analyzed.
3.Model based Collaborative filtering: It is an unsupervised
learning where it mainly concentrates on the characteristics and
Keywords— Movie Recommendation, KNN, Content-based features which mainly gives more impact.
Filtering, Collaborative Filtering, Matrix Factorization, Clustering, Collaborative and content-based information’s are
Hybrid . combined in a recommender system to upgrade
recommendation performance [6].
I. INTRODUCTION
In this world of e-commerce most of the customers prefer II. RELATED WORK
to get products online by spending less time on applying filter The Recommendation system topic is the active platform
on the data. Hence a system which work smartly to suggest for researchers to work always and the recommendations, a
things or items dependent on the past data or information of the person needs are movies, books, web-series, songs, venues,
individual client are every much required. videos, sites and many more. Many Recommendation systems
The Suggestion can play a huge role in business. For are based on the concept of content filtering, memory-based
suppose suggesting a movie, music, web series to a person is filtering and model-based filtering techniques.
depending on various attributes and mainly a personal Recommendation for online system depends on the user
experience such as age, mood, experience, lifestyle can play a input, data collection, ratings, preferences, similarities and
huge role in recommending and these are not generic like other factors. Based on the input the information are collected
grocery items [1]. Using Machine learning for this purpose and then these are used to show the recommendations. The
solves the problem [2]. Hence recommending will be quite a one’s without any input or previous choices the
difficult task since all the factors need to be considered. recommendation is shown based on the ratings and popularity.
However, the Recommendation system provides a huge In today's world a larger group of community indulged in
value in the business. Important aspect of this system is
exploring everything over the internet [7]. [8] author has
previous searched data. However, recommendation systems are
shown how people using the internet exponentially increasing.
the foundation wellspring of these organizations and reduces
the risk to search [3,4]. Filtering techniques that predict preference of users are a
requirement in the current era. These kinds of methodology
The Recommendation system can be classified as three applied in many areas like movies, watching videos, buying
types: Based on content, memory and on collaborative filtering books, news, and buying things from ecommerce, where data
systems. given by user are analysed and best suited are suggested to
user. Using K-means and collaborative approaches the above
purpose is solved [9]. Prateek et.al have done performance

978-1-6654-1960-4/21/$31.00 ©2021 IEEE 769

Authorized licensed use limited to: National Aerospace Laboratories. Downloaded on February 23,2022 at 06:47:35 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV 2021).
IEEE Xplore Part Number: CFP21ONG-ART; 978-0-7381-1183-4

analysis of different kinds of recommendation system. They created and recommended to user irrespective of language or
found that using the user-based collaborative filtering on location [17]. Rupali Hande et.al used user rating matrix and
smaller dataset reduces mean square error. Prince Praveen CF is applied to make recommendations [18]. Netflix dataset
et.al proposed system using content-based approach which used by Yibo et.al for content-based recommendation system
will consider similar items that is the items with similar where result is analyzed, which is showing great improvement
content will be recommended to the user. Also, it uses cosine in recommendation with increase in number of nodes in Spark
similarity algorithms to find similar content [10]. Furtado et.al Clusters [19]. Sudhashu et.al used sentiment analysis by using
used collaborative filtering approach. This will work on user tweets by user for predicting system recommendation [20].
ratings, therefore restricting one’s choice to traverse more. [10,11,12,21,23,25] have done prediction using KNN
This system is based on a collaborative approach. Future work algorithm with cosine similarity, collaborative filtering,
is looking for personal choice such as emotions. Mahmudul clustering, content-based filtering.[22] In this paper the review
Hasan et.al used K-means algorithm for evaluating the rating of approaches and strategies created in the Collaborative
difference between articles and creating a new cluster. They Filtering based are discussed. [24,26] In this paper Neural
also made use of K-medoids clustering algorithm [11]. Prateek network based, rating and sentiment-based filtering system are
Sappadla et.al have made use of the mean squared error discussed.[27] In this paper authors have compared
metrics to calculate the prediction made by their system. In recommendation models like Backpropagation, Singular
this system collaborative filtering and content-based filtering Value Decomposition, Deep Neural Network and Deep Neural
are considered [12]. Network with Trust. And their result implied that DNN with
They have tried out the following basic guidelines approach to trust model outperformed with high accuracy of 83% and with
provide users an idea of the performance from the, Global 0.74 accuracy.
average, the average rating for all users across all movies is
computed. This global average servers as a prediction for all TABLE I. EMPIRICAL COMPARISION ON MOVIE RECOMMENDATION
SYSTEM USING COLLABRATIVE FILTERING
the missing entries in the rating matrix. User average, this
method will evaluate the average rating for each user. This Model Paper Dataset Method Results
technique could be anticipated to execute slightly superior Recom F. MySQL This system progressively
than the global average since it takes into account the rating mendat Furtado, relies on explicit
ion A, Singh individual users' outcomes
behavior of the users. Movie average, in this method, each system (2020) ratings which contrasted with
missing rating entry is given with the average rating for the using will compute different
movie. Adjusted average, they used this method to integrate Collab the connection systems that are
some information about the user and the movie when making orative between based on
filterin Different clients collaborative
a prediction for the entry. This system recommends movies g and relying upon filtering are
even with very small similarity which is the main advantage of KNN each rating, and achieved.
this system. The data of the other users is not required. It algorith prescribes Comparatively
recommends items outside the category of items the user m. movies to Accurate results
people sharing are achieved.
rated. It also recommends new movies which are not rated by common
the users. The crucial challenges they faced while interest.
implementing this system is a scalability issue, that is working Subsequently
with the huge data set resulted in memory constraints. The permitting users
to explore more.
data cannot be stored as a dense matrix due to its enormous Collab Nirav Netflix The system performance is
size. So, they had to make use of sparse matrix representation orative Raval, recommends an measured using
so as for the program to work without any memory issues. filterin Vijayshri item by using system Root
Additionally, intermediate results such as the user - user g Khedkar rating and mean square
approa (2019) similarity
similarity matrix cannot be evaluated and recorded because of ch among the two
the vast memory footprint. using users.
KNN
III. EMPIRICAL COMPARISION OF MOVIE RECOMMENDATION Collab Sang-Min Database to Comparing The genre
SYSTEM orative Choi, store data previous correlations
filterin Sang-Ki method and shortcomings
As Per the Table I and Table II different papers are reviewed g Ko, Yo- revised method, are overcome
for Movie recommendation system where, 2006 George et.al Sub Han the genre here. The
used dataset from Movielens and did prediction based on past (2012) correlation results are
values are analyzed with
evaluated item [13]. Sang et.al improved the algorithm to averaged to respect to two
overcome its disadvantages. Oiming et.al used supervised avoid spurious characteristics
recommendation system [14]. Bo Li et.al proposed a method results of genre
which follows clustering approaches and their performance are correlations.
Collab Bei-Bei Database to Checking for Model based on
evaluated and obtained 84% accuracy [15]. Context based orative CUI store data nearest knn and
algorithm proposed by Gaurav Arora et.al obtaining precise filterin (2017) neighbors and collaborative
results. Manoj Kumar et.al k-means algorithm movie is g giving the filtering.
approa recommendatio

978-1-6654-1960-4/21/$31.00 ©2021 IEEE 770

Authorized licensed use limited to: National Aerospace Laboratories. Downloaded on February 23,2022 at 06:47:35 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV 2021).
IEEE Xplore Part Number: CFP21ONG-ART; 978-0-7381-1183-4

Model Paper Dataset Method Results Model Paper Dataset Method Results
ch ns Omi
using (2019)
KNN content- Muham imdb, NLP Identifies
Collab Qiming IMDb the result is provides good based mad yulp13, and methodology movies &
orative Diao, measured by recommendatio filtering Ibrahim, yulp14 used with categories.
Filterin Minghui McNemar’s test ns by using all algorithms Imran microblog text
g. Qiu, the accessible , Sarwar sentiment.
Matrix Chao- data sources collaborati Bajwa ,
Factori Yuan Wu, ve Riaz Ul-
zation. Alexander filtering, a Amin,
J. Smola, hybrid and
Jing filtering Bakhtiar
Jiang, algorithm Kasi
Chong (2019)
Wang Collaborat Sudhans Tweets were Tweets by This system
(2014) ive hu extracted users are used recommends
filtering, Kumar, from twitter for sentiment movies only
In Table I few papers are listed and compared where Content Shirsend as a analysis for between the
based u S. database. recommendati year 2014-
collaborative filtering is mainly considered for building filtering, Halder, on of movies. 2017, it used
recommendation model. Sang et.al used Database to store data Hybrid Kanjar weighted
where database method used for extraction created cold Filtering De, scored
problems. Here the rating and similarity among the two users and Partha fusion to
sentiment Pratim improve the
are used to develop recommendation system. analysis. Roy. recommenda
(2018) tions, hence
TABLE II. EMPIRICAL COMPARISION ON MOVIE RECOMMENDATION producing
SYSTEM USING HYBRID FILTERING accurate
Method Results results.
Model Paper Dataset
Content- Yibo Netflix All three Increasing
KNN, Sivaku Motion Comparing all Accuracy of based, Wang, dataset System used number of
Decision mar k, pictures, the three Knn-0.6, collaborati Mingmi nodes in
tree, Gopi appraisals, algorithms Logistic-0.7, ve ng Spark
Logistic nath genome_sco dtree-0.7 filtering, Wang,1 clusters, the
Regressio R(2020) re and hybrid and Wei efficiency of
n and unite system Xu algorithms
into one (2018) improved
dataset Recomme Prateek Movielens Global average Various
Cosine Prince Movielens To find similar Hybrid ndation Sappadl serves as Methods
similarity Praveen, content engine system a, Yash prediction for performance
and KNN Praveen between user which means using Sadhwa all missing for 100k
Goud, and making the combining Collaborat ni, entries, User dataset,
Sagar prediction content and ive Pranit average is User-user,
Parmar using KNN collaborative filtering & Arora. incorporated validation=0
(2020) filtering to Content- (2017) into rating .8029,
show the based predictions, Test=0.8366.
recommenda filtering. Movie average K-nearest
tion assigns rating users,
K-means Rishabh Movielens Using utility Root mean to each Validation
and KNN Ahuja, matrix and square missing rating MSE=0.802
Arun choosing the method=1.2 entry movie, 9, Test
Solanki, correct number 3154 Adjusted MSE=0.836
Anand of clusters for No. of average will 6. Latent
Nayyar( classification cluster=68 adjust global factor,
2019) and KNN is average value regularizatio
used to predict for user bias n=0.1,Valida
Clustering Maham Movielens Cluster are Proposed and movie tion
approach udul created by clustering bias. Movie MSE=0.848
is used for Hasan, grouping methods bias is the 6, Test
recommen Farzana similar items improved the difference MSE=0.871
dation Aktar by performance between 0.
Lubna , incrementing in terms of average movie
Sadia the value of mean rating and
Chowdh the similarity absolute global average
ury , matrix by one error. rating. Thus,
Nusrat every time recommendati
Jarin similarity is ons are made.
Haque , found. Content Rupali Web crawler User rating Better
Towsif based Hande, is used to matrix is recommenda
Ahmed filtering, Ajinkya download converted into tions system

978-1-6654-1960-4/21/$31.00 ©2021 IEEE 771

Authorized licensed use limited to: National Aerospace Laboratories. Downloaded on February 23,2022 at 06:47:35 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV 2021).
IEEE Xplore Part Number: CFP21ONG-ART; 978-0-7381-1183-4

Model Paper Dataset Method Results Regression are used but they resulted in overfitting data.
Collaborat Gutti, relevant a pseudo rating developed Hybrid combination of content and collaborative filtering are
ive Kevin movie matrix which hence used with cosine similarity to get high accuracy but this
filtering, shah, content for merges minimizing approach showed cold problem as short come. To improve
Hybrid Jeet the dataset. original rating MAE and performance other factors could be considered like ratings. But
systems. Gandhi, with the increasing
Vrushal processed precision most of the work listed showed sparsity problem. Technique
kamtika result and then and accuracy like clustering, similarity and classification can be used to get
r CF is applied better recommendation thereby minimizing MAE and
(2016) to make increasing performance.
recommendati
ons.
Content Manoj MovieLens This model In light of FUTURE WORK
based Kumar, divided ratings the
filtering, D. K and votes into combined
To get better accuracy and efficient results for the movie
collaborati Yadav, minimum, load of recommendation system, the model-based filtering could be
ve Ankur medium and various considered over content and memory-based filtering and KNN
filtering Singh, maximum attributes are algorithm with matrix factorization could be considered for
and k- Vijay votes. And used for handling the huge platforms and for better accuracy and good
Means Kr. they have used result
algorithm. Gupta ratios evaluation.
recommendations.
(2015) depending on Users across
total number of the globe
votes received, could use
average this system.
REFERENCES
weights to
attributes are [1] G. Adomavicius and A. Tuzhilin, "Toward the next generation of
assigned and recommender systems: a survey of the state-of-the-art and possible
recommendati extensions," in IEEE Transactions on Knowledge and Data Engineering,
ons are made vol. 17, no. 6, pp. 734-749, June 2005, doi: 10.1109/TKDE.2005.99.
to users with [2] I. Portugal, P. Alencar, and D. Cowan, The use of machine learning
highest rating algorithms in recommender systems: A systematic review, Expert Syst.
votes. Appl., vol. 97, pp. 205227, 2018
Content Gaurav MySQL Combining Evaluation
[3] Chen, L. S., Hsu, F. H., Chen, M. C., & Hsu, Y. C. (2008). Developing
based, Arora, CF, content, metrics to
recommender systems with the consideration of product profitability for
Context Ashish Context based measure
sellers. Information Sciences, 178(4), 1032-1048
based, Kumar, algorithm to quality, CF,
Collaborat Gitanjali come up with Content and [4] Bouneffouf, D., Bouzeghoub, A., & Ganarski, A. L. (2013, January).
ive based Sanjay hybrid one to context Risk-aware recommender systems. In Neural Information Processing
algorithm, Devre, recommend algorithms (pp. 57-65). Springer Berlin Heidelberg
Euclidean Prof. movies for e- are used for [5] Adomavicius, G., & Tuzhilin, A. (2011). Context-aware recommender
Distance, Amit users. accuracy systems. In Recommender systems handbook (pp. 217-253). Springer
City Block Ghumar parameters US
distance. e. thus [6] Alvarez, S. A., Ruiz, C., Kawato, T., & Kogel, W. (2011). Neural expert
(2014) providing networks for faster combined collaborative and content-based
precise recommendation.
results. [7] N. Smitha and R. Bharath, "Performance Comparison of Machine
Affinity Bo Li, small Performance is obtained Learning Classifiers for Fake News Detection," 2020 Second
propagatio Yibin number of evaluated 84.71% International Conference on Inventive Research in Computing
n, Liao, exemplars using one of accuracy Applications (ICIRCA), Coimbatore, India, 2020, pp. 696-700, doi:
DBSCAN and the authors 10.1109/ICIRCA48905.2020.9183072.
,hierarchic Zheng owned
[8] https://www.statista.com/statistics/273018/number-of-internet-users-
al Qin website.
worldwide .
clustering (2014)
Hybrid George Movielens Based on Provide [9] Rishabh Ahuja, Arun Solanki, Anand Nayyar, Movie Recommender
methods Lekakos previously useful System Using K-Means Clustering AND K-Nearest Neighbour, January
& evaluated insights . 2019
Petros Items or [10] Prince Praveen, Praveen Goud, Sagar Parmar Movie Recommendation
Caravel features System Approaches International Journal of Engineering Research &
as prediction is Technology (IJERT) Vol. 9 (4),April-2020.
(2006) done. [11] F. Furtado, A, Singh Movie Recommendation System Using Machine
Learning International Journal of Research in Industrial Engineering Int.
CONCLUSION J. Res. Ind. Eng. 9( 1) 84–98, 2020.
In this paper, detailed review is done for Movie [12] Prateek Sappadla , Yash Sadhwani , Pranit Arora Movie Recommender
System , 2017
recommendation system where most of the work used the
[13] Nanou, T., Lekakos, G. & Fouskas, K. The effects of recommendations’
Content and memory-based filtering. But for the huge presentation on persuasion and satisfaction in a movie recommender
platforms it shall be difficult and it shows similar system. Multimedia Systems 16, 219–230 (2010).
recommendations based on the past choices. In the [14] Sang-Min Choi, Sang-Ki Ko, Yo-Sub Han, A movie recommendation
recommendation systems KNN, decision tree, Logistic algorithm based on genre correlations, Expert Systems with
Applications, July 2012

978-1-6654-1960-4/21/$31.00 ©2021 IEEE 772

Authorized licensed use limited to: National Aerospace Laboratories. Downloaded on February 23,2022 at 06:47:35 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV 2021).
IEEE Xplore Part Number: CFP21ONG-ART; 978-0-7381-1183-4

[15] Bo Li, Yibin Liao, and Zheng Qin Precomputed Clustering for Movie [22] Nirav Raval, Vijayshri Khedkar A Review Paper On Collaborative
Recommendation System in Real Time Hindawi Publishing Corporation Filtering Based Movie Recommendation System, International Journal
Journal of Applied Mathematics Volume 2014, Article ID 742341, 9 of Scientific & Technology Research Vol.8 (12),December-2019.
pages, 25 June 2014 [23] Mahamudul Hasan , Farzana Aktar Lubna , Sadia Chowdhury , Nusrat
[16] Gaurav Arora, Ashish Kumar, Gitanjali Sanjay Devre , Prof. Amit Jarin Haque , Towsif Ahmed Omi ACIRS: A Comprehensive Item
Ghumare Movie Recommendation System Based On User’s Similarity Based Clustering Approach to Recommend Appropriate Items in
International Journal of Computer Science and Mobile Computing Recommender System, 2019
IJCSMC, Vol. 3(4), April 2014, pg.765 – 770 [24] Muhammad Ibrahim, Imran Sarwar Bajwa , Riaz Ul-Amin, and Bakhtiar
[17] R. Manoj Kumar, D.K. Yadav, Ankur Singh, Vijay Kr. Gupta A Movie Kasi, A Neural Network-Inspired Approach for Improved and True
Recommender System: MOVREC International Journal of Computer Movie Recommendations, Hindawi Computational Intelligence and
Applications (0975 – 8887) 124 (3), August 2015 Neuroscience Volume 2019, Article ID 4589060, 19 pages- 4 August
[18] Rupali Hande, Ajinkya Gutti, Kevin Shah, Jeet Gandhi, Vrushal 2019
Kamtikar Asst. Professor,- Moviemender-A Movie Recommendation [25] Bei-Bei CUI, Design and Implementation of Movie Recommendation
System International Journal Of Engineering Sciences & Research System Based on Knn Collaborative Filtering Algorithm ITM Web
[Gutti* et al., 5(11): November 2016] Conference 12, 04008 (2017)
[19] Yibo Wang, Mingming Wang, and Wei Xu A Sentiment-Enhanced [26] Qiming Diao, Minghui Qiu, Chao-Yuan Wu, Alexander J. Smola, Jing
Hybrid Recommender System for Movie Recommendation: A Big Data Jiang, Chong Wang Jointly Modeling Aspects, Ratings and Sentiments
Analytics Framework,22 March 2018 for Movie Recommendation (JMARS) , August 2014
[20] Sudhanshu Kumar ,Shirsendu S,Halder ,Kanjar De ,Partha Pratim Roy [27] Choudhury, S.S., Mohanty, S.N. & Jagadev, A.K. Multimodal trust
Movie Recommendation System using Sentiment Analysis from based recommender system with machine learning approaches for movie
Microblogging Data arXiv:1811.10804v1, 27 Nov 2018 recommendation. Int. j. inf. tecnol. (2021).
https://doi.org/10.1007/s41870-020-00553-2
[21] Sivakumar K, Gopinath R, Machine Learning Algorithms for Movie
Recommendation, Journal of Advances in Electronics and
Multidisciplinary computer Science engineering Vol:1, Issue:1, , pp:29-
37,Dec 2002.

978-1-6654-1960-4/21/$31.00 ©2021 IEEE 773

Authorized licensed use limited to: National Aerospace Laboratories. Downloaded on February 23,2022 at 06:47:35 UTC from IEEE Xplore. Restrictions apply.