Solving Cold Start problem in Trust-Based Recommendation Systems Using Temporal Pagerank Algorithm

Collaborative filtering is a extensively used method for recommendation systems. The main assumption is that users with comparable taste in the past are expected to give similar ratings on the items of interest in the future. However collaborative filtering naturally suffers from Cold Start phenomenon. This problem presented by cold-start users who rate zero or a few items. As a result, there has been developing research in recent years to provide effective recommendation for cold start user. In this paper, we propose a method to solve cold start new user problem that accommodate user model withtrust and distrust model with consideration of friendship time factor to find trusty users as reference users. The suggestions of these users then collected to provide appropriate recommendation. Experiments based on Epinions data-set demonstrate that our method outperforms other counterparts both in coverage and user coverage rate

Keywords:

Collaborative-filtering , Cold-start , Coverage , Recommendation-system , Trusty-users