Hybrid Recommendation Algorithm based on User Behavior
Shucheng Guo, Chen Li · 2020
Due to the large amount and strong sparsity of data of e-commerce users' behaviors, the traditional single recommendation algorithm cannot meet the requirement of the recommendation system for accuracy. This paper proposes a hybrid recommendation algorithm (KMFSCF) that integrates K-means and Funk-SVD into the collaborative filtering algorithm. Use K-means to cluster users, and then decompose the scoring matrix of user-item after clustering, calculate user similarity to make scoring prediction, and finally make recommendations according to the ranking of ratings. Experiments show that this algorithm can effectively reduce the influence of data sparsity, reduce the recommendation error and improve the recommendation accuracy.