Design of Hybrid Recommendation Algorithm Based on User Dynamic Behavior and Static Attributes
Xiaoxiao Dai · Journal of Physics Conference Series · 2021
Abstract Collaborative filtering recommendation algorithm is one of the most successful recommendation algorithms, but the traditional collaborative filtering recommendation algorithm is proved to have a series of problems in data sparsity, cold boot and scalability. Based on the problems mentioned above, this paper analyses user dynamic behavior and a hybrid recommendation algorithm based on user dynamic behavior and static attributes (UDBSA) was proposed in this paper. By determining the optimal value of BP critical point for the number of ratings, the recommendation strategy was dynamically selected according to the number of ratings. This method can comprehensively alleviate the influence of problems of new users and concept drift related problem on recommendation results.