Personalized Hybrid Educational Recommendation Framework Using Machine Learning Techniques

Abstract

Recommendation systems are among the most important online information services in several domains, including the educational field. This study aims to provide personalized recommendations by building a framework for a personalized educational recommendation system using a hybrid approach that integrates collaborative filtering and content-based filtering using Self-Organizing Map (SOM) clustering algorithm along with the Term frequency – Inverse Document Frequency technique (TF-IDF) and cosine similarity algorithm. A weighted fusion hybridization with switching hybridization were used to combine the collaborative filtering approach and the content-based approach. Using switching and weighting techniques to combine recommendation technologies addresses the cold start problem and scalability issues. It is observed that the best results are achieved when selecting a weight of 0.8 for content and 0.2 for collaborative filtering. The use of the SOM algorithm in the hybrid system demonstrated the best results, achieving a recall rate of 77% while the precision rate was 23.93%. Also, the results showed an MAE of 0.0038 and an RMSE of 0.0759 that is the lowest error rate compared to the other algorithms.