Research on User Behavior Analysis and Personalized Recommendation Based on CatBoost Algorithm
Download as PDF
DOI: 10.25236/icceme.2023.012
Corresponding Author
Shiyang Li
Abstract
Considering the problems of data sparsity, cold start, and classification feature processing in user behavior analysis and personalized recommendation in Internet applications, this paper proposes a research method for user behavior analysis and personalized recommendation based on the CatBoost algorithm. First, based on the user's historical behavior data, a user profile and interest model are established to obtain information such as the user's primary attributes, behavioral characteristics, interest preferences, and consumption capabilities. Second, in order to eliminate the prediction bias in the gradient-boosting decision tree, an ordered boosting strategy is proposed. Thirdly, based on the classification feature processing capability of the CatBoost algorithm, a personalized recommendation model that comprehensively considers the influence of user features and item features is constructed to describe the matching degree between users and items accurately. Finally, a user behavior analysis and personalized recommendation system based on the CatBoost algorithm is built. In addition, the effectiveness of the method proposed in this paper is proved through simulation and experimental verification.
Keywords
User behavior analysis; Personalized recommendation; CatBoost algorithm; Classification feature processing; Ordered boosting