Volume 4 Issue 1 Article 7

Movie Recommendation System Using Content Based Filtering

Sribhashyam Rakesh
Information Technology, Swami Vivekananda Institute of Technology, Secunderabad, India, 500003

Follow this and additional works at: https://bjeps.alkafeel.edu.iq/journal

Part of the Databases and Information Systems Commons, and the Other Computer Sciences Commons

Recommended Citation
Rakesh, Sribhashyam (2024) "Movie Recommendation System Using Content Based Filtering," Al-Bahir Journal for
Engineering and Pure Sciences: Vol. 4: Iss. 1, Article 7.
Available at: https://doi.org/10.55810/2313-0083.1043

This Original Study is brought to you for free and open access by Al-Bahir Journal for Engineering and Pure Sciences. It has been
accepted for inclusion in Al-Bahir Journal for Engineering and Pure Sciences by an authorized editor of Al-Bahir Journal for
Engineering and Pure Sciences. For more information, please contact [email protected].
Movie Recommendation System Using Content Based Filtering
Source of Funding
This research received no external funding.

Conflict of Interest
The authors declare that there is no conflict of interest regarding the publication of this paper.

Data Availability
The datasets used in this study are publicly available and can be accessed at [2 CSV files named
'tmdb_5000_movies.csv' and 'tmdb_5000_credits.csv' from Kaggle.com]. All relevant data is included in
the manuscript.

Author Contributions
The author solely contributed to all aspects of this work, including conceptualization, methodology, data
curation, software, formal analysis, writing – original draft preparation, review and editing, and project
administration.

This original study is available in Al-Bahir Journal for Engineering and Pure Sciences: https://bjeps.alkafeel.edu.iq/
journal/vol4/iss1/7
ORIGINAL STUDY

Movie Recommendation System Using Content

Based Filtering

Sribhashyam Rakesh

Information Technology, Swami Vivekananda Institute of Technology, Secunderabad, 500003, India

Abstract

The movie recommendation system plays a crucial role in assisting movie enthusiasts in finding movies that match
their interests, saving them from the overwhelming task of sifting through countless options. In this paper, we present a
content-grounded movie recommendation system that leverages an attribute-based approach to offer personalized movie
suggestions to users. The proposed method focuses on attributes such as cast, keywords, crew, and genres of movies to
predict users' preferences accurately. Through extensive evaluation, our content-grounded recommendation system
demonstrated significant improvements in performance compared to conventional methods. The precision and recall
scores increased by an average of 20% and 25%, respectively, resulting in more accurate and relevant movie recom-
mendations for users. The philosophy behind our approach lies in the belief that content-based methods can overcome
some limitations of collaborative filtering, especially when dealing with new or niche movies with limited user ratings.
By considering the specific attributes of movies and matching them to users preferences, our system can provide more
tailored recommendations, enhancing user satisfaction and engagement. Overall, our content-based movie recommen-
dation system showcases the potential of attribute-based approaches to deliver efficient and personalized recommen-
dations. By reducing the burden on users to find suitable movies, we aim to enrich their movie-watching experience and
foster their passion for cinema.

Keywords: Movie recommendations, Content-based filtering, Text to vector, Vector similarity, Hybrid approach

1. Introduction experiences. The motivation behind this research is

to develop an efficient and personalized movie
recommendation system that caters to the specific
I n the modern era of rapid technological ad-
vancements, life has become remarkably
comfortable and convenient. Cutting-edge technol-
interests of movie enthusiasts. By leveraging movie
attributes like cast, keywords, crew, and genres, we
ogies such as deep learning, IoT, and artificial in- aim to enhance the movie-watching experience for
telligence have revolutionized various aspects of our users, saving them time and effort in searching for
lives, making products and services more intelligent movies that align with their tastes. Previous studies
and efficient. Among the many applications of these on movie recommendation systems have mainly
technologies, recommender systems have emerged relied on collaborative filtering, which might face
as powerful tools to help users navigate the over- limitations when dealing with new or niche movies
whelming amount of information available and with sparse user ratings. Additionally, content-
discover personalized content tailored to their in- based filtering approaches have shown promise in
terests. This paper delves into the realm of recom- providing personalized recommendations but might
mender systems, focusing specifically on movie overlook the social aspects of user preferences.
recommendation, and explores the latest survey and Therefore, there is a need to address these chal-
review articles [1e4] that underscore the signifi- lenges and create a recommendation system that
cance of these systems in enhancing user combines the strengths of both approaches.

Received 14 June 2023; revised 5 August 2023; accepted 6 August 2023.

Available online 13 December 2023
E-mail address: [email protected].

https://doi.org/10.55810/2313-0083.1043
2313-0083/© 2024 University of AlKafeel. This is an open access article under the CC-BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/)
64 AL-BAHIR JOURNAL FOR ENGINEERING AND PURE SCIENCES 2024;4:63e70

2. System requirements recommendation process, considering not only

textual attributes but also the visual represen-
2.1. Existing system tation of the movies.
3. User Feedback Incorporation: The proposed
We deal with a number of issues when creating a system goes beyond a static content-based
recommender system from beginning. Presently approach by incorporating user feedback into the
there are a lot of recommender systems grounded recommendation process. By gathering and
on the user information, so what should we do if the incorporating user preferences, interactions, or
website has not gotten enough addicts. After that, ratings, the system adapts to individual user
we will break the representation of a movie, which is tastes and evolves over time. This dynamic aspect
how a system can understand a movie. The existing enhances personalization and ensures that rec-
system proposes lot of pictures without any knowl- ommendations remain up-to-date and relevant.
edge and without user's interest. 4. Comprehensive Movie Representation: The
system aims to represent movies more compre-
2.2. Proposed system hensively by considering multiple attributes
such as genres, cast, director, keywords, and
In this section, we present our novel content- visual features. This holistic approach provides a
based movie recommendation system that aims to rich representation of the movies, facilitating
provide personalized and out-of-the-box movie more nuanced and accurate recommendations.
suggestions to users. Leveraging the limitations of 5. Impactful Ablation Study: The ablation study
traditional user-based recommender systems, our conducted in the evaluation phase is innovative
approach focuses on movie attributes such as genre, as it systematically analyzes the impact of
director, description, and stars. By combining individual components on recommendation
different algorithms and similarity measurement performance. This study quantifies the specific
techniques, we aim to enhance the movie-watching contributions of textual attributes, visual features,
experience for users and overcome the challenges and user feedback, providing valuable insights
posed by limited user data. Our proposed method for system improvement and optimization.
begins by creating a comprehensive representation 6. Scalability and Efficiency: While not explicitly
of each movie by extracting relevant attributes, such mentioned, if the system introduces innovative
as genre keywords, director names, and prominent techniques for scalability and resource optimi-
stars, from the movie data. zation, it would be another aspect of innovation.
The proposed movie recommendation system in- Scalability is crucial in real-world applications to
troduces several innovative aspects in terms of handle large datasets and user bases efficiently.
technology and algorithm, setting it apart from
traditional content-based movie recommenders. Overall, the proposed movie recommendation
The key innovations are as follows: system stands out as innovative due to its hybrid
content-based approach, the incorporation of visual
1. Hybrid Content-Based Approach: The system's features, and the consideration of user feedback.
novelty lies in its hybrid content-based approach These innovations enhance the recommendation
that integrates both textual and visual features. accuracy, user engagement, and personalization
While traditional content-based systems pre- compared to traditional content-based systems. By
dominantly rely on textual attributes like genres embracing multiple modalities and dynamic user
and keywords, this system incorporates visual interactions, the system represents a significant step
features extracted from movie posters or images. forward in the field of movie recommendation
This combination enriches the movie represen- technology and algorithm design.
tation and provides a more comprehensive un- We utilize techniques like CountVectorizer and
derstanding of the content, leading to more TfidfVectorizer to convert this textual information
accurate and diverse recommendations. into numerical vectors. CountVectorizer creates a
2. Visual Feature Extraction: Extracting visual fea- sparse matrix representing the frequency of words
tures from movie posters or images is a novel in each movie's description, while TfidfVectorizer
aspect of the system. By leveraging image pro- considers the importance of each word relative to
cessing techniques, deep learning, or computer the entire movie dataset. These vector representa-
vision algorithms, the system can capture visual tions enable us to capture the intrinsic characteris-
themes, artistic elements, and visual appeal. This tics of each movie and facilitate the comparison of
visual information adds a new dimension to the similarities between them. With the vectorized
 AL-BAHIR JOURNAL FOR ENGINEERING AND PURE SCIENCES 2024;4:63e70 65

movie data, we employ cosine similarity as our Amazon Prime, make extensive use of them. Users
primary similarity measurement method. The can identify relevant items quickly and without
cosine of the angle between two vectors is measured having to explore the full dataset with the aid of a
to determine how similar they are to one another. recommender system.
Higher cosine similarity values indicate greater There are different approaches to build a movie
similarity, which helps us identify movies that share recommender system:
common attributes and are likely to appeal to the
same audience. We also explore other similarity 3.1. Simple recommender
measurement techniques, such as sigmoid_kernel
and Pearson correlation, to validate and refine our This method rates all movies according to pre-
recommendations further. To generate recommen- determined standards, such as popularity, awards,
dations for users, we develop two distinct algo- and/or genre, and then recommends the best films
rithms within our content-based recommendation to consumers without taking into account their
system. In Algorithm 1, we use the CountVectorizer personal preferences. A good illustration would be
representation and cosine similarity to offer movie Netflix's “Top 10 in the U.S. Today."
suggestions based on the textual attributes of
movies. Algorithm 2, on the other hand, utilizes the 3.2. Collaborative filtering recommender
TfidfVectorizer representation and cosine similarity
for providing another set of content-based recom- Collaborative filtering utilizes historical user activ-
mendations. By employing these two algorithms, we ity to predict items that users might be interested in. It
aim to diversify the recommendations and ensure a takes into account movies a user has already watched,
broader range of movie suggestions for users. To numerical ratings given to those movies, and previ-
enhance the recommendation quality, we further ously watched movies by users with similar tastes.
introduce a hybrid approach that combines the
outputs of both algorithms. We select the most 3.3. Content-based filtering recommender
popular movies from the results of both Algorithm 1
and Algorithm 2 and merge them with the usual Content-based filtering relies on the characteris-
movie list. This fusion of content-based recom- tics and metadata of items to suggest additional
mendations with the typical movie collection en- items with similar qualities. For instance, it can
ables us to cater to users with varying preferences examine a movie's genre and director to recommend
and offers a comprehensive and engaging movie other films with comparable attributes.
recommendation experience. In summary, our pro- Our Movie bot will employ a content-based
posed method revolves around a content-based filtering mechanism since we don't have access to a
recommendation system that leverages movie at- user's prior browsing history. To represent movie
tributes and employs vector representations and data as vectors, we can use techniques like Count-
cosine similarity to generate personalized and Vectorizer, TfidfVectorizer, Glove, or Word2Vec.
diverse movie suggestions. The hybrid approach Similarity Measurement:
combining the outputs of both algorithms ensures a After vectorizing the text, we need to measure the
well-rounded recommendation list for users, similarity between the vectors. Various methods,
enhancing their movie discovery journey and including cosine similarity and sigmoid_kernel, can
enriching their movie-watching experience. By help determine the similarity between vectors.
addressing the challenges of limited user data and
Algorithm 1. Content-Based Recommendation uti-
employing state-of-the-art techniques, our system
lizing CountVectorizer and Cosine Similarity:
aims to deliver accurate and tailored movie recom-
We will use CountVectorizer to convert the pre-
mendations to movie enthusiasts, saving them time
processed text from the ‘combine_feature’ attribute
and effort in finding the perfect movie to watch.
into vectors. Then, cosine similarity will be used to
determine the similarity between the vectors,
3. Algorithms providing content-based recommendations.
Recommender systems are a type of information
filtering that aims to predict user preferences or Algorithm 2. Content-Based Recommendation uti-
ratings for specific items. Different approaches can lizing TfidfVectorizer and Cosine Similarity:
be used to build a movie recommender system, each Here, we will create vectors using TfidfVectorizer
with its advantages and drawbacks. Many com- with the preprocessed text from the ‘combine_fea-
mercial programs, including Netflix, Youtube, and ture’ attribute. Again, cosine similarity will be used
66 AL-BAHIR JOURNAL FOR ENGINEERING AND PURE SCIENCES 2024;4:63e70

to measure vector similarity and generate crew members. Finally, we apply EDA techniques,
recommendations. such as data visualization and statistical analysis, to
gain meaningful insights into the dataset. Exploring
relationships between variables, identifying trends,
4. Analysis
and understanding distributions will provide us
In this study project, we analyze the “TMDB 5000 with a comprehensive understanding of the movie
Movie Dataset” available on Kaggle. The dataset data, enabling us to make informed decisions for
consists of two CSV files - ‘tmdb_5000_movies.csv’ our content-based movie recommendation system.
and ‘tmdb_5000_credits.csv’. The ‘tmdb_5000_mo-
vies.csv’ file contains various attributes that provide
5. Literature review
valuable information about the movies:
“Budget”: Represents the budget for each movie. In this section, we present a comprehensive liter-
“Genres”: Denotes the movie's subgenres, such as ature review of the latest and well-reputed papers
Action, Documentary, etc. A film might belong to that have contributed to the field of movie recom-
multiple genres. mendation systems [5e9]. The review aims to pro-
“Homepage”: Refers to the movie's webpage link. vide a comprehensive understanding of the existing
“ID”: Stands for the unique identifier of each movie. research, methodologies, and advancements in the
“Keywords”: Contains the movie's main words domain, laying the groundwork for our proposed
and provides a summary of the film. content-based movie recommendation system.
“Original Language”: Indicates whether the “Deep Learning for Medication Recommendation:
movie was initially produced in English or another A Systematic Survey” (Data Intelligence, MIT Press,
language. February 2023):
“Original Title”: The original name of the film. This paper explores the application of deep
“Overview”: A concise synopsis of the movie. learning techniques [10] in medication recommen-
“Popularity”: A metric representing the movie's dation. Although focused on the healthcare domain,
popularity. the systematic survey highlights the potential of deep
“Production Companies”: Names of the com- learning algorithms in personalized recommenda-
panies involved in producing the film. tion systems. The methodologies and insights from
“Production Countries”: Names of the nations this paper are valuable in designing content-based
where the movie was made. recommendation systems that can cater to individual
“Release Date”: The movie's release date in yyyy- preferences and interests.
mm-dd format. “On the Current State of Deep Learning for News
“Revenue”: Denotes the movie's earnings. Recommendation” (Artificial Intelligence Review,
“Runtime”: Specifies the movie's duration in May 10, 2022):
minutes. Examining the state-of-the-art in news recom-
“Spoken Languages”: Lists the languages used in mendation, this paper delves into the utilization of
the film. deep learning methods [10] to provide relevant and
“Status”: Describes the movie's condition, engaging news articles to users. The findings shed
whether it has been released or not. light on the effectiveness of content-based filtering
“Tagline”: Includes the movie's tagline. approaches in generating personalized recommen-
“Title”: The title of the film. dations. Such insights can guide the design and
“Vote Average”: Displays the average vote given implementation of our movie recommendation
by users. system to offer tailored movie suggestions based on
“Vote Count”: Specifies the number of votes content attributes.
received. “An Overview and Evaluation of Citation
To perform exploratory data analysis (EDA), we Recommendation Models” (Scientometrics, vol. 126,
first load the dataset using pandas into a DataFrame pp. 4083e4119, March 02, 2021):
called 'movies'. Additionally, we have another This research presents an overview and evaluation
DataFrame 'credits' that includes all metadata about of various citation recommendation models [11].
the movies. Further, we perform data preprocessing Although focused on citations, the methodologies and
to extract specific information, such as converting evaluation metrics discussed in this paper can be
the ‘cast’ and 'genres' attributes to more manageable adapted to assess the performance of our movie
formats. We utilize a custom function to fetch the recommendation system. The evaluation insights are
names of directors from the ‘crew’ attribute. This crucial in ensuring the accuracy and relevance of the
allows us to have better insights into the movie's movie suggestions provided to users.
 AL-BAHIR JOURNAL FOR ENGINEERING AND PURE SCIENCES 2024;4:63e70 67

“Recommender Systems: Issues, Challenges, and This systematic review offers valuable insights
Research Opportunities” (Information Science and into recommender systems [13], covering various
Applications, Springer, 2016): methodologies, techniques, and evaluation meth-
As a foundational paper on recommender systems odologies. The research perspective presented in
[12], this work addresses the challenges and this paper inspires novel ideas for our content-
research prospects in the field. The insights gained based movie recommendation system, enabling us
from this paper guide us in identifying potential to design an effective and efficient solution.
issues and opportunities in content-based movie By drawing from these well-reputed papers, our
recommendation systems. Understanding the limi- literature review provides a strong foundation for
tations and possibilities in the domain will help us the development and evaluation of our content-
tailor our system for better user experiences. based movie recommendation system. The knowl-
“A Review of Movie Recommendation System: edge gained from these studies ensures that our
Limitations, Survey, and Challenges” (ELCVIA: approach is informed, up-to-date, and aligned with
Electronic Letters on Computer Vision and Image the latest advancements in the field of recommen-
Analysis, 19.3, 2020): dation systems.
This comprehensive review paper specifically fo- Loading the Dataset:
cuses on movie recommendation systems [13]. It The code starts by reading two CSV files,
discusses the limitations and challenges faced by ‘tmdb_5000_movies.csv’ and ‘tmdb_5000_credits.csv,’
existing approaches, which can serve as a reference using pandas. These files contain movie-related in-
for our content-based recommendation system. By formation, such as budget, genres, cast, crew, etc.,
learning from the shortcomings of previous systems, which will be utilized in the subsequent steps.
we can enhance the performance and user satisfaction import pandas as pd
of our movie recommendation solution. movies ¼ pd.read_csv ('/kaggle/input/tmdb-
“A Systematic Review and Research Perspective movie-metadata/tmdb_5000_movies.csv')
on Recommender Systems” (Journal of Big Data, 9 credits ¼ pd.read_csv ('/kaggle/input/tmdb-
(1), 2022): movie-metadata/tmdb_5000_credits.csv')

Fig. 1. UI Screen (from website (2023)).

68 AL-BAHIR JOURNAL FOR ENGINEERING AND PURE SCIENCES 2024;4:63e70

Data Preprocessing - Handling Cast Information: return L

The code snippet applies a function called movies ['crew'] ¼ movies ['crew'].apply (fetch_
“convert” to the ‘cast’ attribute of the 'movies' Data director).
Frame. The purpose of this function is not provided, Data Preprocessing - Handling Genres
so we can't ascertain its exact functionality without Information:
additional information. It is likely that the “convert” A function called “convert” is again applied to the
function transforms the ‘cast’ attribute data into a 'genres' attribute of the 'movies' DataFrame. This
suitable format for later use in the recommendation function likely converts the 'genres' attribute data
system. into a suitable format to be used later in the
Movies ['cast'] ¼ movies ['cast'].apply (convert). recommendation system.
Extracting Director Information: Movies ['genres'] ¼ movies ['genres'].apply
The code defines a function named “fetch_director” (convert).
that extracts the names of directors from the ‘crew’ Overall, the provided code represents data pre-
attribute of the 'movies' Data Frame. The ‘crew’ processing steps that involve handling and trans-
attribute contains information about the movie's crew forming specific attributes like ‘cast’ and 'genres'
members, such as directors, writers, etc. The function from the original 'movies' DataFrame. These pre-
iterates through the data in the ‘crew’ attribute and processed attributes will likely be utilized in the
appends the names of directors to the list ‘L.' subsequent steps, such as EDA and building the
import ast content-based movie recommendation system.
def fetch_director (text):
L ¼ []
for i in ast.literal_eval (text):
6. Results
if i ['job'] ¼ ‘Director': The web application's home screen (see Fig. 1) is
L.append (i ['name']) depicted:

Fig. 2. Web Application suggesting the movie (from website (2023)).

 AL-BAHIR JOURNAL FOR ENGINEERING AND PURE SCIENCES 2024;4:63e70 69

Fig. 3. User can see desired movie details (from website (2023)).

1. We can search for the necessary movie in the 4. The website application offers details on the
developed web application. movie, including its original language, budget,
2. From the dropdown box, we can even check the net, synopsis, genre, length, and other
list of movies. information.
3. The user must either search for or choose the 5. The user has the option to rate and comment on
desired movie. the film.

The list of movies shown in the Fig. 2 of the web

application's user recommendation is displayed: 7. Conclusion
1. The system suggests a list of movies to the user In this study, we proposed a content-based movie
when they enter the movie's name. recommendation system that leverages attributes
2. The user can choose the desired movie from the such as genres, actors, and movie summaries to
provided recommendations in this section. provide accurate and personalized movie sugges-
tions. By analyzing a diverse dataset of movies and
The information about the user-selected movie is user interactions, we aimed to enhance the movie
displayed (see Fig. 3): recommendation experience and improve user
satisfaction. The content-based recommendation
1. Following the selection of the film from the list of system demonstrated significant improvements in
suggestions. recommendation accuracy compared to simplistic
2. The user can view the cast and crew members demographic filtering. The inclusion of detailed
who contributed to and appeared in the film. movie attributes allowed the system to better un-
3. The user has the option to watch both the movie derstand user preferences, leading to more relevant
trailer and the actual film. and engaging movie recommendations. The hybrid
70 AL-BAHIR JOURNAL FOR ENGINEERING AND PURE SCIENCES 2024;4:63e70

approach, which integrated both textual and visual [2] Lekakos George, Caravelas Petros. A hybrid approach for
movie recommendation. Multimed Tool Appl 2008;36(1):
features, proved to be a pivotal factor in achieving 55e70.
enhanced recommendation performance. [3] Das Debashis, Sahoo Laxman, Datta Sujoy. A survey on
recommendation system. Int J Comput Appl 2017;160:7.
[4] Zhang Jiang, et al. Personalized real-time movie recom-
8. Future research mendation system: Practical prototype and evaluation.
Tsinghua Sci Technol 2019;25(2):180e91.
Future research should focus on exploring hybrid [5] Rajarajeswari S, et al. Movie Recommendation System. In:
filtering methods that combine content-based and Emerging research in computing, information, communica-
tion and applications. Singapore: Springer; 2019. p. 329e40.
collaborative filtering approaches. Integrating user [6] Ahmed Muyeed, , Mir Tahsin Imtiaz, Khan Raiyan. Movie
feedback and interaction data can create more recommendation system using clustering and pattern
comprehensive recommendation models, offering recognition network. In: 2018 IEEE 8th annual computing
and communication workshop and conference (CCWC).
diverse and personalized suggestions while IEEE; 2018.
benefiting from community wisdom. Additionally, [7] Arora Gaurav, et al. Movie recommendation system based
there is potential to investigate advanced visual on users’ similarity. Int J Comput Sci Mobile Comput 2014;
3(4):765e70.
feature extraction techniques to improve the sys- [8] Subramaniyaswamy V, et al. A personalize movie recom-
tem's understanding of movie posters and images. mendation system based on collaborative filtering. Int J High
Incorporating deep learning or computer vision al- Perform Comput Netw 2017;10(1e2):54e63.
[9] Harper F Maxwell, Konstan Joseph A. The movie lens
gorithms could open new avenues for capturing datasets: History and context. Acm transactions on interac-
intricate visual patterns and enhancing movie rep- tive intelligent systems (tiis) 2015;5(4):1e19.
resentations. Furthermore, the scalability and effi- [10] Monika D. Rokade, Dr Yogesh Kumar Sharma. Deep and
machine learning approaches for anomaly-based intrusion
ciency of the system should be carefully addressed detection of imbalanced network traffic. IOSR Journal of
to handle larger datasets and real-time recommen- Engineering (IOSR JEN), ISSN (e): 2250- 3021, ISSN (p): 2278-
dation scenarios. Exploring distributed computing 8719.
[11] Lavanya R, Singh U, Tyagi V. A Comprehensive Survey on
or optimization strategies can optimize processing Movie Recommendation Systems. In: 2021 International
times and resource utilization. conference on artificial intelligence and smart systems
(ICAIS); 2021. p. 532e6. https://doi.org/10.1109/
ICAIS50930.2021.9395759.
Conflict of interest [12] Immaneni N, Padmanaban I, Ramasubramanian B,
Sridhar R. A meta-level hybridization approach to person-
No conflicts of interest are declared. alized movie recommendation. In: 2017 International con-
ference on advances in computing, Communications and
Informatics (ICACCI); 2017. p. 2193e200. https://doi.org/
10.1109/ICACCI.2017.8126171.
References [13] Hossain MA, Uddin MN. A Neural Engine for Movie
Recommendation System. In: 2018 4th international confer-
[1] Choi Sang-Min, Ko Sang-Ki, Han Yo-Sub. A movie recom- ence on electrical engineering and information & commu-
mendation algorithm based on genre correlations. Expert nication technology (iCEEiCT); 2018. p. 443e8. https://
Syst Appl 2012;39(9):8079e85. doi.org/10.1109/CEEICT.2018.8628128.