One Hidden Semantic Model Based on Intergroup Effects for E-Commerce

Abstract

E-commerce systems often collect data that clearly express user preferences without considering the remaining negative cases, which gives rise to the hidden semantic problem. In this paper, we improve the original hidden semantic model and propose an intergroup effect model that incorporates users’ historical browsing behavior, user type, and browsing content; by adopting the weighting and add weighting factors, we can predict users’ preferences for different products more accurately and match the candidate products with users’ current behaviors, so as to give more reasonable and effective product recommendation results; by adding the group effect model of user group and product group, we can achieve more accurate prediction of user preferences and make the recommendation more reasonable and effective. The research shows that the hidden semantic method based on intergroup effects information is better than other basic methods at a certain identified evaluation stage. In practice, users’ purchasing preferences change with time, and using a hidden semantic method based on intergroup effects recommendation can effectively improve the recommendation quality of e-commerce recommendation systems.