9 II February 2021

https://doi.org/10.22214/ijraset.2021.33043
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

Automated Online Course Recommendation

System using Collaborative Filtering
Ashuta Thakur1, Sheetal Rathi2
1, 2
Department of Computer Engineering, Thakur College of Engineering & Technology, Mumbai, India

Abstract: In the digital world of online information huge amounts of Massive Open Online Courses (MOOCs) are available of
different category and domain. Multiple online courses are available on different platform finding appropriate course from this
massive available course is difficult for students. Recommender system plays vital role in finding appropriate courses to students.
Managing massive amount of information and identifying individual users’ choice and behavior has become tedious task
nowadays, so the aim of recommender system is to suggest relevant course to student based on user behavior and similarity with
another course. Several recommender system techniques are being implemented like content based, collaborative, Knowledge
based. This paper aims to build a hybrid approach using collaborative filtering with content base filtering. This system
recommendation is based on course description and ratings. Experiments were conducted on real datasets to get the overall
performance of proposed system.
Keywords: Recommendation system, Collaborative Filtering (CF), Content based Filtering (CBF), Knowledge based Filtering,
MOOCs.

I. INTRODUCTION
With the ever-growing large volume of online information, recommender systems are an efficient strategy to beat such information
overload. Recommender systems are the systems that are designed to recommend things to the user supported various factors.
Companies like YouTube, Netflix, Amazon, etc. use recommender systems to help their users to recommend the correct product,
video or movies for them. Advances in technology has changed the way of education. Massive Open Online Courses (MOOCs) are
capable of providing several learners to access courses over the web. Recommender System (RS) is computerized system that
suggest/recommend item to user.
The number of MOOCs and the number of students registered in MOOCs are growing per annum. In 2018, more than 900
universities were offering MOOCs with 11,400 courses available, and around 101 million students had registered in them (Shah,
2018), providing learners with a good sort of choices. With such a high number of courses available, learners now face the matter of
choosing courses without being overwhelmed.
With the rise in e-commerce and online business, the number of users interested in online Web services has increased. Both MOOC
providers and online businesses advertise their courses and services while learners look for courses that match their interests and
needs. In these situations, recommender systems play a crucial role, and have attracted the attention of researchers. Recommender
systems are algorithms and techniques that, based on their preferences, suggest matching and related courses or services to the
learner, knowledge about which comes from learner profiles and systems-gathered histories. Recommender systems help MOOC
providers grow and learners find more appropriate and customized services based on their personalities and interests.
Recommender systems discover patterns in considerable datasets to find out preferences of different users and predict items that
correlate to their needs. Recommender systems is divided into two broad categories: collaborative filtering recommender systems
and content-based recommender systems. Collaborative filtering recommender systems perform recommendations on users who
have had similar taste in the past will make similar choices in the future. Content based recommender systems consider the profile
of users and items.
The online course recommendation systems suggest to the students the best courses in which they are interested. This paper presents
a recommendation methodology that recommends courses to students based on similarity between courses taken by the target
student and other students. It aims to provide an effective course recommendation using multiple techniques. The students will be
clustered into groups based on traditional data-mining (DM) techniques will to Collaborative filtering using knowledgebase.

©IJRASET: All Rights are Reserved 222

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

II. RECOMMNDATION TECHNIQUES

Recommendation System is primarily arranged in three sections dependent on the available data. Available data assumes significant
part in Recommendation System. Three principal Approaches of Recommendation System are Content based Recommendation
System, Collaborative Filtering Recommendation System, and Knowledge Based Recommendation System

Recommendation
Technique

Content-Based Collaborative Knowledge Based

Filtering Filtering Filtering
Technique Technique Technique

Figure1: Types of Recommendation Technique

A. Content-based Filtering Technique
Content based filtering is based on information filtering examine users’ profile in detail for analyzing data. CBF utilize descriptive
keywords associated with each item to make recommendations Content based contains information about a user data and choice of
user. Mainly, two type of data is required one is users’ profile and other is user’s interest. The taste is based on how user rates the
items. In this CBF technique, rated items will be compare not rated items and similarity between rated items will be evaluated for
recommendation. Items that are almost similar to the completely rated ones, will be recommended to the user. Amazon is a best
example of content-based filtering. Accuracy of content-based filtering is dependent on user’s data. There are two approaches to
content-based filtering
1) Recommendation using User Rating Approach: In this approach product will be rated on user’s choice and similar product
rating will be predicted. Advantage of this approach is it works without user reviews.
2) Recommendation used User Description: In this type of approach full description if item is required. Detail description of item
like title, type, keyword, price, tag line, genre etc is required. This text format strings are converted to number for calculating
similarity. Term Frequency-Inverse Document Frequency (TF-IDF) is used in extraction features of information.

Drawbacks:
a) Problem of New-user -Same as for collaborative filtering, user profiles are required as input.
b) Limited analysis of content -The recommender highly depends on the information available from the documents. Therefore, the
documents contain either some machine-readable text or they need to be classified by users manually. The popularity of two
documents, which has the same vectors, cannot be differentiated by such content-based systems.
c) Over specialization-The recommender recommends the items similar to the ones and the one which is already known, leading
to a portfolio effect.

B. Collaborative filtering Technique

Collaborative filtering Approach is one of most usually utilized algorithm. This recommender system suggests things dependent on
c similarity measures among items and user. The things prescribed to user are liked by comparative User. These comparable user
structure a gathering called neighborhood. Assume a user has not evaluated thing suggestion will be given dependent on
comparative taste user rating. Collaborative filtering technique has two classification user-based and item based. User based is
similitude between the user’s choice and item based is comparability among item and different items.

A few downsides of this strategy are as per the following:

1) Problem of New User - For suggesting things, the user should indicate their preferences first without this data, no
recommendations can be made.
2) Problem of New-Items – Items that are new in the system need ratings by users before they are being utilized in the
recommendation process

©IJRASET: All Rights are Reserved 223

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

Collaborative
Filtering
Technique

Model Based Memory

Filtering Based Filtering
Technique Technique

User Based Item Based

Figure2: Types of Collaborative Filtering Technique

C. Knowledge-based filtering Technique

Knowledge based recommender system depends on domain knowledge and about the user’s knowledge. Removing the knowledge
of learner’s and knowledge about the available information’s, is the significant task in knowledge-based recommender system.
Knowledge-based recommender systems won't consider assembling long -term generalizations about their users however they favor
creating a recommendation supported matching between user’s want, preferences and available set of items. This methodology
doesn't have the issue of over specialization and sparsity because in this methodology each user is independent with another user
and the statistical evidence. Indeed, this methodology is sensitive to change in the learner interest and preferences of learner and
furthermore it doesn’t have any dependency on information rating. Knowledge-based methodology need not have an initial database
of learner’s preference and also it is capable to exploit the data concerning the training domain to produce the most effective answer
to the learners. However, the point of this methodology is to give the most appropriate recommendations and reasoning about how
learning materials of the domain meets the learner’s need.

III. RELATED WORK

Many research works have conducted in the field of online course recommender system, content-based filtering, collaborative
filtering, data mining and machine learning methods. MOOCs are increasing in huge amount and need for recommender system has
also increases already many applications uses recommendation techniques. Applications such as Amazon, Netflix, Facebook,
LinkedIn, Twitter etc. uses Recommender system with different approach. We provide a brief overview of the previous research on
the development recommendation based on machine learning models in this section.
Amer Al-Badarenah [1] proposed system for course selection based on collaborative filtering technique. Various recommendation
system techniques like content-based filtering, classification method, collaborative method, and knowledge-based method where
compared. In proposed system recommendation of course is done on basis of similar users. It recommends user course and expected
grades for course. Association rule mining technique is applied to discover comparable. The proposed method demonstrated in the
experiment demonstrates that association rule mining is a desirable tool for making recommendations. To verify the overall
performance of the proposed method, experiments were conducted with real data sets. The downside of the Automated
Recommender Method for Course Selection is that it takes a very long time to mine student rules and low performance at the same
time.
Sheetal Girase[7], adopted hybrid approach od content-based approach and collaborative filtering approach to recommend colleges
to students. College recommending system dataset is collected from 16 IITs, 30 NITs, 20 Government Funded Institutes, and 60
Maharashtra State Engineering Colleges. To decrease the efficiency of algorithm matrix factorization (MF) using Singular Value
Decomposition (SVD)is used. System faces big challenge in ranking data which assures user satisfaction. Cold Start is one of the
common problems with the recommendation method. As in recommendation system, lot of quality information is needed, but there
is situation where there is lack of information. Cold start issue happens when there are new users to the system, new products or a
completely new system where there is no data on them, so it is difficult to recommend the recommended system.
Obeidat, R [4] Apriori Algorithm Association Rule Mining and SPADE Algorithm were used in proposed system. Course
recommendation is done on three parameter the coverage, the support and the confidence factor of the rule.

©IJRASET: All Rights are Reserved 224

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

Recommending courses for large dataset is difficult as Association rules have high confidence and low coverage on the dataset
hence, one low coverage rule will affect the average coverage of the rules.
M. Mohamed [2] The definition of ranking is most common in the recommendation system. Therefore, in the case of a rating-based
recommendation system, this data sparsity issue would have this problem, since certain users will score the subset of the overall
database because most common products would have less ratings that will impact the accuracy of the recommendation system.
Scalability: The system needs more resources for processing information and forming suggestions for users with the increase in
users and objects. The majority of tools available are used to classify users with common interests and similar tastes and
descriptions. The combination of various types of available techniques and physical enhancement of structures solves this form of
problem. Some computations are also calculated offline to improve the speed of the recommendations online.
Over-specialization: Over-specialization is another problem in the recommendation system, especially in content-based filtering.
The primary goal of such systems is to suggest items that are strongly compatible with consumer preferences; however, this could
lead to indicating items that have already been visited by the user. For example, in the news recommendation system, which is based
on content-based filtering, a user can receive recommended news that she has already read and that the consumer does not like
different ones.
Fang Liu [11] Apriori algorithm is introduced in core function design. Designed scheme is applied in one agricultural major
category of China Open University system to find student course selection frequent pattern and strong association rules in Animal
husbandry and veterinary medicine major category.
User Based Collaborative Filtering (UBCF) approach builds similarity between users by M. Kommineni, P [1]. Recommendation of
books using similarity measures proved to be efficient using user based collaborative filtering. Similarity methods were done using
PCC, CPCC, Cosine, Jaccard where CPCC (constarined pearson correlation) similarity measure is the best measure for calculating
similarities among users.
D. Estrela [6] Proposed hybrid approach by combining Content based filtering and Collaborative filtering approach to recommend
online course for elearners. Hybrid approach faces cold start problem and precision of recommender was not accurate.
JIANG Ya-tong [12] A hybrid recommendation system combining the features of collaborative, content based for personalized
subject recommendation for students. Each subject score is compared with each student using collaborative filtering. Pearson
correlation and angle cosine are used to calculate similarity is proposed system.
Hai-hui Wang [8] Modern learning style MOOCs, massive open online course, provide online courses to unlimited users and open
access via the web. It becomes difficult for the students to decide which course to opt within limited time. Jaccard’s Similarity is
used to cluster data.
V. Manvitha[13]The goal of the proposed system was to solve the issue of cold starts. It is a recommendation system that advises
users to play music. The Taste Profile dataset obtained from the Million Song Dataset is the dataset used. 10M is the dataset scale.
Moreover, no person relies on online sites to buy songs, apparels, books, rented films, etc. Competition occurred between different
online sites that forced owners of the Websites that provide their users with personalized music recommendations. So, the
recommender systems came into existence. The issue with cold starting is that the recommenders do not recommend entirely new
users or products for those users or items for which adequate knowledge is not available. The combination of association rules and
clustering technique is an algorithm based on this music recommender.
Effective Recommender system method was also tried during the research, the most popular would be [14], from Amazon.com.
They use recommendation algorithms to personalize the online store for each customer. Rather than matching the user to similar
customers, item-to-item collaborative filtering matches each of the user’s purchased and rated items to similar items, then combines
those similar items into a recommendation list. item-to-item collaborative filtering
algorithm is scalable over very large customer and product catalogs, requires very less processing time to recommend product.
Tanay Kulkarni [3] In proposed system student learning process is being monitor based on available resources so that student’s area
of interest can been known to recommend. Huge amount of information available on the web so getting relevant information which
fulfills user’s searching criteria is a main concern. Therefore, to address this issue the primary step of model is to develop a keyword
focused web crawler which extracts relevant URLs that includes the keyword searched by the user and considers only such web
pages as significant for further processing, thus increasing the accuracy and search efficacy. The system will get URLs for user
activity analysis and other users will receive recommendations. The subject of focus in this paper is the use of techniques for content
fetching and effective user profiling, so that the system gives out accurate recommendations.

©IJRASET: All Rights are Reserved 225

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

M Viswa Murali [5] Proposed system recommender model “UBCF (User Based Collaborative Filtering) model” generates the
recommendation from the closest user behavior using the whole matrix saved. Internally the model calculates the cosine similarity
among all consumers represented as vectors.
Kiratijuta Bhumichitr[9]Choosing university courses is complicated, so this method focuses on advising students to take elective
courses. Since many elective open courses are available, but students do not have time to explore, this system often allows students
to know the correct course to be chosen. It recommends a course based on enrollment data from students.This paper is used by most
well-known algorithms to discover comparable students who use Pearson Correlation Coefficient and Alternating Least Square
Correlation Coefficient to collaboratively filter (ALS). The experimental results show that with 86 percent accuracy, applying ALS
in this domain is superior to collaborative.Student enrollment information was not adequate to assess student conduct to support
better recommendation. Downside of this method.

IV. THE PROPOSED RECOMMENDATION SYSTEM

There has recently been an increase in demand for online courses and distance education. Instead of institutes, students tend to study
online, so the key explanation for the online course recommendation system is to recommend relevant courses to students. To boast of
their skills and profession, students learn from MOOC.They consider factors such as course length, current market patterns for
courses, faculty teaching for specific courses, catalogs, interest, etc., to choose the appropriate course.
The proposed system focuses to use collaborative filtering technique with knowledge-based to recommend best appropriate course to
students. On available dataset the proposed system will map all similar students based on K means algorithm. After Collaborative
filtering technique will be hybrid with knowledge-based to recommend appropriate course to students.

Following is the proposed system flow:

Figure3: Proposed system model flow diagram

1) Step 1: Data Collection and Preprocessing

Data available is very less and insufficient to build a recommender system. Many MOOCs don t provide confidential dataset of their
users and courses. This proposed system requires ratings of courses so web scrapping is the technique through which this data was
pulled. Web Scraping is a method used to retrieve vast quantities of website data by downloading the data and saving it to a local
file on your machine or to a database in table (.csv) format. Data scraping is a effective method to extract data from websites (based
on the website s regulations). Dataset of proposed system was of Udemy which is one of best MOOCs platform.
Data available through scrapping is a raw data which needs to be transform in a useful format.Raw Data is cleansed through
processes such as filling in missing values or removing incomplete rows of data, smoothing out noisy data, or solving data
inconsistencies. Irrelevant data is being dropped from raw data. The cleaned dataset contains live and available public courses.

©IJRASET: All Rights are Reserved 226

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

Name of dataset | df_courses | df_reviews

-------------------- | ----------------| -----------
Number of rows originally | 10.000 | 1.415.734
Number of rows after cleaning | 8.834 | 1.391.194
Final number of columns | 30 | 4

2) Step 2: Data Analysis and Visualization

Data analysis is the heart of the machine learning based projects. After preprocessing of the data next step is to analyze and
investigate data sets and summarize their main characteristics by data visualization methods. Data analysis answers the various
questions such as what data can tell us before building system, how much data is sufficient? How to establish the relationship
between learned data and result. Data analysis help us to actually predict top courses in MOOCs, total subscribers, number of
courses which can be helpful for clustering.
Data Analysis helps us to
· to offer insight into a knowledge set.
· Understand the data structure.
· Extract important relationships that hold between various parameter.
It is an honest practice to know the info first and check out to collect as many insights from it.

3) Step 3: Clustering
Aiming to identify similarity in course description and course objective, in this step, course is clustered using the K-means
algorithm. K-means clustering is one of the simplest and popular unsupervised machines learning algorithms.
Clustering refers to the task of classifying a set of objects into a set of homogeneous groups such that the objects within the same
group (i.e., cluster) are most similar while having the greatest dissimilarity to objects in other groups. The K-means algorithm works
by randomly selecting points as the initial centroids of clusters.
A measure of distance (e.g., Euclidean distance) is then calculated for each one of the other points, and each one is assigned to the
cluster having the closest centroid. Subsequently, a new centroid is computed for each one of the clusters. This iterative process of
assigning points to clusters and updating the centroids continues until the sum of squared errors is minimized. Clusters chosen by
proposed system is K=8. Each cluster are label by top 5 words that helps to understand which cluster contain which courses
computed for each one of the clusters.
Moreover, PCA is conducted with three, four, and five components for the purpose of dimension reduction.

4) Step 4: Recommendation System

Proposed system tries to focus on Collaborative filtering with knowledge base. Collaborative filtering is based on user preferences
whereas knowledge base is dependent on additional knowledge about user from profile and course information.
In proposed system course reviews where not appropriate for collaborative filtering. To resolve this after clustering step additional
knowledge of courses features where used with collaborative filtering.

In a knowledge-based algorithm, that includes feature matrix includes:

1) Additional features to recommender system
2) Scaled Feature Matrix
3) Similarity Measure

New features were added to recommender system according to categories in clusters. Different features have different magnitude to
scale for example rating of courses varies from 0 to 5 using scikit-learn library features where standardizes. Similarity measures
between different courses is done by cosine similarity. Cosine Similarity measures angle between two vectors in multidimension
space here two vector refers to addition features of courses. Proposed system recommends course of user’s previous course if any or
similar course based on current selected course id.

©IJRASET: All Rights are Reserved 227

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

V. RESULTS
We performed experiments using courses dataset taken by Udemy. Top 10 courses records are shown Figure4. Around 900 courses
where there without reviews/ratings. The average time period of a course is 26 months. Machine learning course had highest number
of subscribers.Clustering algorithm based on the course descriptions showed good results. Multiple k-s number of clusters were tried
out in kMeans algorithm. For optimal number of clusters k=8 shown in figure5.MAEcalculates the average magnitude of the errors
without considering their path.

r^ui: ratings predicted

rui: ratings known
The deviation between predicted rating and the actual rating of different users and items are measured using the above formula.
Lower MAE values indicate higher accuracy. Overall prediction accuracy of item based and user-based algorithms is very high as
MAE values for all possible neighborhood size falls in the range of 0.33 to 0.38. From the observation of figure 6 item based and
user-based algorithm results, there is not much difference in accuracy but there is lots of difference in proposed algorithm.

Figure4: Top 10 courses

Figure5: kMean parameter

©IJRASET: All Rights are Reserved 228

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at www.ijraset.com

Figure6: MAE

VI. CONCLUSION
Course recommendation system help the students in course choosing process. Classical recommendation techniques like content-
based filtering, collaborative filtering, and knowledge-based filtering are used in different commercial websites. For recommending
course to students’ similarity is performed for course topics and keywords. Proposed system recommends course to students based
on users rating and course description. This online course recommender system focused on hybrid approach of collaborative and
content-based filtering to recommend courses to students. The similarity measures between the user and courses are precise by
making use of the cosine similarity technique. The user is suggested with the top-rated courses according to his interest.
In the future, we can use other information of user so that it would determine the behavior of the student using NLP for further
recommendation.
REFERENCES
[1] Madhuri Kommineni , P.Alekhya, T. Mohana Vyshnavi, V.Aparna, K Swetha, V Mounika,“ Machine Learning based Efficient Recommendation System for
Book Selection using User based Collaborative Filtering Algorithm” IEEEAugust 2020.
[2] M. Mohamed, M. Khafagy and M. Ibrahim, "Recommender Systems Challenges and Solutions Survey", 2019 International Conference on Innovative Trends in
Computer Engineering (ITCE), 2019.
[3] Tanay Kulkarni, Madhur Kabra , Dr. Radha Shankarmani ,“ User Profiling Based Recommendation System For E-Learning” IEEE2019.
[4] Obeidat, R., Duwairi, R., Al-Aiad, A. ,“ A Collaborative Recommendation System for Online Courses Recommendations ” IEEE2019.
[5] M Viswa Murali, Vishnu T G, Nancy Victor ,” A Collaborative Filtering based Recommender System for Suggesting New Trends in Any Domain of Research
” International Conference on Advanced Computing & Communication Systems (ICACCS) 2019.
[6] Estrela, D., Batista, S., MartinhoD., & Marreiros, G., “A Recommendation System for Online Courses”, Springer International Publishing AG 2017.
[7] Sheetal Girase, Varsha Powar and Debajyotimukhopadhyay, “A User-friendly College Recommending System Using User-profiling And Matrix
Factorization Technique”,International Conference on Computing, Communication and Automation, 10.1109/CCAA.2017.8229779.
[8] Hai-hui Wang, Chalothon Chootong, Ankhtuya Ochirbat, Worapot Sommool, W K T M Gunarathn, and TimothyK.Shih, “Online Courses Recommendation
System Based on Industry Occupation Skills Requirements”, IEEE, 2017.
[9] Kiratijuta Bhumichitr, Songsak Channarukul, Nattachai Saejiem, Rachsuda Jiamthapthaksin and Kwankamol Nongpong, “Recommender Systems for
University Elective Course Recommendation”, IEEE, 2017.
[10] Amer Al-Badarenah and Jamal Alsakran, “An Automated Recommender System for Course Selection”, International Journal of Advanced Computer Science
and Applications, vol.07, pp. 3, 2016.
[11] Fang Liu, Shaogang Zhang, Junyi Ge, Feng Lu and Jun Zou, “Agricultural Major Courses Recommendation Using Apriori Algorithm Applied in China Open
University System”, International Symposium on Computational Intelligence and Design, 10.1109/ISCID.2016.1109.
[12] JIANG Ya-tong，fu Qiang，li Fei，lv Hai-xia，wu Gang, and Mehmet Fatih Usl, “Personalized Recommendation and Analysis Method for Student Partiality
for One or Some Subject(s) In HigherEducationManagement”, IEEE, 10.1109/ICMSE.2014.6930479.
[13] V. Manvitha, M. Sunitha Reddy “Music Recommendation System Using Association Rule Mining and Clustering Technique to Address Cold start Problem”,
International journal of engineering and computer science,Vol. 3, No. 6855-6858, 2014.
[14] Linden G, Smith B, York J. Amazon.com recommendations: Item- to- item collaborative filtering[J]. IEEE Internet Computing, 2003, 7(1): 2003: 76- 80.
[15] Xiwang Yang, Chao Liang, and Miao Zhao, “Collaborative Filtering-Based Recommendation of Online Social Voting”, IEEE Transactions on Computational
Social Systems,10.1109/TCSS.2017.2665122.
[16] Dhruv Shah, Pratik Shah and Asim Banerjee, “Similarity Based Regularization for Online Matrix-factorization Problem: An Application to Course
Recommender Systems”, IEEE, 10.1109/TENCON.2017.8228164.
[17] Suleyman Uslu, Can Ozturan, and Mehmet Fatih Usl, “Course Scheduler and Recommendation System for Students”, IEEE, 10.1109/ICAICT.2016.7991812.
[18] Skills Requirements”, IEEE, 10.1109/UMEDIA.2017.8074083.
[19] Kaustubh Kulkarni, Keshav Wagh, Swapnil Badgujar and Jijnasa Patil, “A Study of Recommender Systems with Hybrid Collaborative Filtering”,International
Research Journal of Engineering and Technology (IRJET), Volume: 03 ,2016.
[20] Yiu-kai Ng and Jane Linn, “Crsrecs: A Personalized Course Recommendation System for College Students”, International Journal of Advanced Computer
Science and Applications, 10.1109/IISA.2017.8316368.
[21] S. Ray and A. Sharma, "A Collaborative Filtering Based Approach for Recommending Elective Courses", Information Intelligence, Systems, Technology and
Management, pp. 330-339, 2011.

©IJRASET: All Rights are Reserved 229