Leveraging User Personality Information And Item Tags For One Class Collaborative Filtering

Personalized recommendation system has become the important tool of e-commerce companies in order to meet customer demand.However,online users continuously expand and produce large amounts of information,how to use this information and accurately provide personalized content and service is becoming more and more challenging.Traditional recommended models rely too much on the user's preference score,and often affected by data sparseness.One class collaborative filtering(OCCF)model is more suitable for the electronic commerce but there are some obvious shortcoming in item recommendation due to much missing records and extreme imbalance of category,such as data sparsity and cold start problem.In order to deal with the challenge of the above,many researchers devotes themselves to identify the positive and negative case in data,and this article uses the user's personality information and items tag's information as complementary resources in wAMAN model.We proposes two major weighting strategies of negative case: One is embed user personality to calculate user's preference,and the other is to unite user personality and item tag to calculate user's preference.The basic idea of weighting strategies is the more similar between the user and item,the less weight value we should assign to that negative case.The experimental verification on LastFM dataset results show that the proposed model can improve the performance of traditional collaborative filtering model and effective in relieving the data sparsity and new user cold start problem.