Designing Trust Propagation Algorithms based on Simple Multiplicative Strategy for Social Networks

Abstract

In the recent years, with the huge popularity of Web based Social Networks, the trust and trust related issues become more and more important. Many trust metrics have been designed in literature for calculating propagated trust in social networks. They are either of type global or local. Global trust metrics build a global rank or trust value for the nodes or users in the network. On the other hand, local trust metrics take source node as the input and calculate the personalized trust value between that node and a destination node by trust propagation mechanisms. In this paper, we have designed three trust propagation algorithms based on simple multiplicative strategy of trust propagation by taking path length and decay of direct trust values along the trust path into consideration in calculating propagated trust. The accuracy of the algorithms in predicting propagated trust values are calculated and compared with the mole trust [7] algorithm. We have used Mean absolute Error (MAE) as our evaluation metric to measure the extent of deviation of the propagated trust values from the actual trust ratings. We have also examined whether there is any strong correlation between direct trust and propagated trust values calculated by our algorithms.