Trust Inference Using Implicit Influence for Item Recommendation

Abstract

Trust plays a very important role in many existing ecommerce recommendation applications. Social or trust network among users provides an additional information along with the ratings for improving the user reliability on the recommendation. However, in real world, trust data is sparse in nature. So, many algorithms are built for inferring trust. In this paper, we propose a new trust inference method based on the implicit influence information available in the existing trust network. This approach uses the transitivity property of the trust for trust propagation and scale-free complex network property to limit the propagation length in the network. In this regard, we define a new terminology, degree of trustworthiness for a user, which adds the global influence in the inferred trust. This process improves the recommendation accuracy from the existing trust-based recommendation and neighborhood-based collaborative filtering. Due to the availability of users preference from trust network which is absent in rating data, it also alleviates the very well-known cold start users problem of a recommender system. We evaluate the proposed approach on two established real world datasets and report the obtained results.