A Hybrid Recommender System to Enrollment for Elective Subjects in Engineering Students using Classification Algorithms

• Published in: International journal of advanced computer science & applications Vol. 11; no. 7
• Main Author: Rivera, Jerson Erick Herrera
• Format: Journal Article
• Language: English
• Published: West Yorkshire Science and Information (SAI) Organization Limited 2020
• Subjects: Algorithms; Colleges & universities; Engineering education; Hybrid systems; Machine learning; Natural language processing; Recommender systems; Students
• ISSN: 2158-107X; 2156-5570; 2156-5570
• DOI: 10.14569/IJACSA.2020.0110752

One of the main problems that engineering university students face is making the correct decision regarding the lines of elective subjects to enroll based on available information (preferences, syllabus, schedules, subject content, possible academic performance, teacher, curriculum, and others). Under these circumstances, this research work seeks to develop a Hybrid Recommender System. For this, a model based on the Content-based approach of all the subjects that has been studied is developed (using Natural Language Processing and the statistical measures Term Frequency and Inverse Term Frequency), giving it appropriate relevance with the grades that the student has achieved. In addition, a model based on a Collaborative Filtering approach is developed, establishing relationships between different students, identifying similar academic behaviors. Thus, the system will recommend to the student in which lines of elective subjects to enroll to obtain better results in the academic field. The given recommendation will be obtained from machine learning models (XGBoost and k-NN) based on the similarity between the contents of each subject with respect to the line of elective subject and based on the academic relationship between all the students. To achieve the objective, data from engineering students between 2011 and 2016 has been analyzed. The results obtained indicate that the recommendations reach a MAP-k of 82.14% and a precision of 91.83%.