New Graph Based Trust Similarity Measure

Abstract

Trust network in social networks can be considered as graph which trustors and trustees are graph vertices and edges present trust between them with measured values. To evaluate trust between trustors and trustees there is some similarity measures to measure similarity between trustors together or trustees together and then by using evaluated values predict trust value between them. Similarity measure has important effect on final accuracy. In this paper we propose graph based similarity measure. Similarity between two users is computed by connection between them on graph then this computed similarity used with k- nearest neighbors method to evaluate(predict) trust between users. To the best of our knowledge this is the first work introduces graph based similarity measure, empirical results on two real datasets show accuracy of predicted trust using proposed similarity measure outperforms accuracy of method without it.