Social Recommendation Algorithm Based on Learning Interests and Styles for Online Blended Learning Platform

Information overload is a common problem in the construction of online platforms for blended learning, and personalized recommendation is a solution to this challenge. However, the traditional collaborative filtering recommendation algorithm is not fully applicable to online blended learning platforms characterized by sparse rating data and rich text and social information. In this paper, a new model of user learning interest similarity measurement was constructed using the author-topic model based on user-generated text and behavior information. Comprehensive user trust relationships were constructed and comprehensive trust was calculated based on user interaction behavior information by taking trust and distrust factors into full account. User learning styles were automatically identified by analyzing users’ multidimensional learning behavior patterns. Finally, a social recommendation algorithm integrating interest similarity, comprehensive trust, and learning styles was proposed. In addition, an experimental analysis was conducted using the actual data set of the online blended learning platform China University MOOC. The results indicated that the recommended method in this paper can better recommend learning resources of interest to users and effectively improve the recommendation accuracy, thus improving users’ learning effect.