Towards an Efficient Machine Learning Algorithm for a Graduate Study Elective Course Recommendation System

The choice Master’s students have to make regarding elective subject selection in a chosen specialization is really decisive. A wrong decision may affect their personal and academic goals and may impact negatively on their future professional direction. Making bad choices on the elective subjects to offer at this stage may lead to loss of interest by the student, which can result to dropping out of the higher degree program. It is therefore important that students are given support so as to make the right choices regarding elective courses in a chosen specialization, using decision support systems. There are records of successful use of recommender systems to suggest items to users in several domains, like education, e-commerce, entertainment domains, and the like. In this work, ۵ supervised machine learning algorithms are evaluated to determine the most efficient for the training and prediction on elective courses to be offered by a Master’s Degree student based on the student’s background knowledge of undergraduate courses, and on the academic record of previous Master’s students. This research employed an experimental approach, and a Python software application is modeled with the Object-Oriented Analysis and Design (OOAD) using the standards notations and techniques of the Unified Modelling Language (UML) to build the recommender system. The result from the evaluation of the five machine learning models shows that the Naïve Bayes and Decision Tree algorithms have equal accuracy value of ۹۹.۴۰۹%, which is the highest and an equal F۱ Score of ۰.۹۱۸, which is equally the highest. Decision tree was selected to be the classifier model for the recommender system.

کلیدواژه ها:

Educational Data mining ، Recommender system ، machine learning algorithms