Dynamic Credibility Threshold Assignment In Trust And Reputation Mechanisms Using Pid Controller

An e-marketplace is an example of a multi-agent system where buyers try to find the best seller with best Quality of Service (QoS). The uncertainty of open marketplaces have resulted in the design of reputation systems that help buyers find honest feedback from their peers (advisers). Despite the advances in this field, there is no systematic approach for setting the honesty threshold as an acceptable level of honesty of advisers in the Trust and Reputation Management (TRM) systems. Having an appropriate honesty threshold is important in these systems, since having a high threshold would filter away possibly helpful advisers, or the opposite - having a low value for it may permit malicious advisers to badmouth good services. In this paper we propose a self-adaptive honesty threshold management mechanism that adopts PID feedback controller from the field of control systems. Experimental results on a real-world dataset show that having a dynamic honesty threshold increases the successful transaction rate of buyers in a marketplace, and improves the accuracy of the TRM system used in that marketplace.