Trust inference algorithms for social networks

Abstract

The exponential growth of social networks has made establishing a trusted relationship increasingly important. Recommender systems can play an important role in assessing a user’s trustworthiness. Such systems are designed to offer recommendations of trustworthiness when establishing connections among social network members, where the system rates members by inferring their degrees of trust. In this work, we developed a recommender system that provides recommendations about trusted social network members. We compared the time complexity and the accuracy of the following four adapted algorithms and a new proposed algorithm: Top Trusted Members, Target’s Reputation and Similarity, Depth First Search (DFS) Trust Propagation, Dijkstra’s Trust Propagation and Target’s Followers (new). An experiment was conducted using a dataset from Twitter. The results show that the Target’s Followers algorithm is a promising approach for making accurate recommendations, especially when the network is dense.