A Learning Resource Recommendation Algorithm Incorporating User Information and Rating Differences

Abstract

Despite the growing popularity and development of large-scale open online learning platforms, they have been suffering from the problem of “information disorientation.” To increase students' learning efficiency, it is important to build recommendation algorithms based on students' basic information and historical rating data. In this research, we propose a new collaborative filtering recommendation algorithm that incorporates the user Information and the rating differences. The algorithm first uses the user information labels to calculate the user similarity, then introduces rating differences to enhance the conventional cosine similarity based on the characteristics of non-preferred rating data, and finally linearly combines the two similarities. The experimental results demonstrate that the algorithm enhances the recommendation effect of learning resources. The MAE and RMSE is employed to quantify the prediction accuracy of the recommendation algorithm.