Social Recommendation Based on Multi-relational Analysis

Abstract

Social recommendation methods, often taking only one kind of relationship in social network into consideration, still faces the data sparsity and cold-start user problems. This paper presents a novel recommendation method based on multi-relational analysis: first, combine different relation networks by applying optimal linear regression analysis, and then, based on the optimal network combination, put forward a recommendation algorithm combined with multi-relational social network. The experimental results on Epinions dataset indicate that, compared with existing algorithms, can effectively alleviate data sparsity as well as cold-start issues, and achieve better performance.