A Collaborative Filtering Algorithm Based on the User Characteristics and Time Windows

Abstract

Collaborative filtering algorithm is a widely used recommendation algorithm. In the traditional collaborative filtering algorithm, a single user similarity calculation method is usually considered, and the user’s own attribute characteristics are not used as the basis of neighbor user selection. At the same time, in the process of recommendation, user’s interest is considered to be static and given the same weight in different time periods, without thinking the dynamic changes of user’s interest. For above problems, this paper proposes a collaborative filtering algorithm based on the user characteristics and time windows. Firstly, a collaborative filtering algorithm based on item rating and user’s own attribute characteristics is proposed in the process of calculating similarity. Secondly, the dynamic time windows are divided according to the Ebbinghaus forgetting curve to reflect the user’s short-term interests in the recommendation process, the concept of time function is added to assign different time weights to user interests in different periods in the process of interest fusion. Finally, through experimental analysis, the recommended effect of the algorithm is significantly improved compared with the traditional collaborative filtering recommendation algorithm.