A Novel Recommendation Method Based on User's Interest and Heterogeneous Information

It's a consensus that trust relationship is significant to improve the recommendation efficiently. But in most cases, trust relationship information is so sparse and difficult to use. Actually, the trust relationship is the response of interest among users, that is, it is an effective method to find the appropriate trust relationships by mining users' interests accurately. There are so many factors that can affect users' interest as well, such as age, occupation and so on. Based on these factors we can construct a heterogeneous information network, this paper deeply mine more accurate trust relationship through the interest and similarity from the heterogeneous information network among users, and merges the trust relationship to the matrix decomposition techniques. Moreover, we innovative conduct our experiment to test the recommendation algorithm based on trust, which has not been studied so far in MovieLens100k dataset. Experimental results demonstrate that our method outperforms other counterparts both in terms of accuracy.