Hybrid Recommendation Algorithm Based on Trust Relationship and User Preference

摘要

Most current trust-based recommendation algorithms are only based on reviews of trust neighbors, and do not consider users' preference. To handle this limitation, we propose a hybrid recommendation algorithm in social network (HRSN). Our proposed method constructs a trust network based on explicit trust relationships of a target user, and then identifies an opinion leader from his/her trust neighbors who can affect the target user and give proper comments on items. The proposed method integrates the comments of opinion leaders and user preferences to predict ratings on target. To demonstrate the effects of the proposed method, we conduct experiments on Epinions dataset by comparing with TrustMF, SoReg, SoRec and SVD++. The results show that HRSN can achieve the lowest MAE, and outperforms SVD++, TrustMF, SoReg and SoRec 11.6%, 9.6%, 7.5% and 1.7% respectively, which suggests that HRSN can provide a better recommendation than other compared methods.