Inferring trust using relation extraction in heterogeneous social networks

Abstract

People use trust to cope with uncertainty which is a result of the free will of others. Previous approaches for inferring trust have focused on homogeneous relationships and attempted to infer missing information based on existing information in a single relationship. In this paper we propose using methods of social network analysis to infer trust in a heterogeneous social network. We have extended the problem of relation extraction and allowed using any type of binary operator on matrixes, whereas previous work have focused on linear combination of base matrixes (the only allowed operator was summation of two matrixes). We present two genetic algorithms which use ordinary numerical and fuzzy logic operators and compare them on a real world dataset. We have verified our claim - ability to infer trust in a heterogeneous social network- using proposed methods on a web-based social network. © 2008 Springer-Verlag.