A Courses Recommendation System based on Graph Clustering and Ant Colony Optimization in MOOC Environment

Abstract

Course recommendation is a tricky issue for many students. When they just have a limited information about the courses, they must rely on general course schedule systems for guidance, but the outcome are inadequate. In this research, we propose a course recommender system based on graph clustering and ant colony optimization to solve the problem. Graph representation, graph clustering, user weighting using ant colony optimization, and rating prediction are the four essential steps in the proposed method. The dataset is first represented as a weighted network based on the degree of similarity between each pair of users. The Pearson-r correlation coefficient is used to calculate the similarity between users. Ratings of users who are most similar to the active user are the goal of the clustering phase. As a result, users in the same cluster have a lot of similarities. Through third step, the Ant colony algorithm is used to weight users (students) with the goal of picking a group of highly correlated users associated with their importance values as target user's neighbor users. Finally, unknown ratings are predicted in the fourth phase by combining the rating values of nearby users with their weights of similarity to the target user.