Personalized Recommendation Method Based On User Hierarchical Structure

With the development of e-commerce,recommender system has become a tool to assist users to select goods.At the same time,it is increasingly difficult for recommender system to accurately recommend the user preferred items in a large number of items.Traditional recommender methods rely on users' rating data to recommend items.However,in ecommerce scenarios,the explicit feedback data is difficult to obtain.The implicit feedback data that users click on or buy items has gradually become the focus of attention.At present,one class collaborative filtering model is suitable for processing implicit feedback data,but there are still cold start problems.Meanwhile,it's hard to effectively distinguish between positive and negative cases in implicit feedback data.In order to solve these problems,this paper uses user's hierarchical structure information as auxiliary information,dividing hierarchical structure into explicit hierarchical structure and implicit hierarchical structure according to user's autonomous choice and intrinsic attributes.Combining with one class collaborative filtering model,we propose one class collaborative filtering method with user hierarchical structure(UHS_OCCF)and one class collaborative filtering method with weighted user hierarchical structure(WUHS_OCCF).User groups and personality information in the public Mypersonality dataset are used to construct user explicit and implicit hierarchical structures respectively.The experimental results show that the proposed two methods based on user hierarchical structure can effectively improve the accuracy of recommendation and alleviate the cold start problems.