Diffuse-like recommender with enhanced similarity of objects

Abstract – In last decades, diversity and accuracy have been regarded as two essential measures in evaluating recommender systems. However, a clear concern is that a model focusing excessively on one measure will put the other one at risk, which means that it is not easy to greatly improve diversity and accuracy simultaneously. In this paper, we propose a method which can be used in many being diffuse-like algorithms, like ProbS, BHC and HHP, to get more relevant and personalized recommendation results on user-object bipartite networks. The main idea is to enhance the RA similarity of objects in the transfer equations, by giving a tuneable exponent on the shoulder of RA similarity, to intentionally regulate the resource allocation on objects of different degrees. Experiments on three benchmark data sets, MovieLens, Netflix, and RYM show that the modified algorithms can yield remarkable performance improvement compared with the original models.