An Adaptive Collaborative Filtering Algorithm for Online Reputation Systems

This paper presents an adaptive collaborative filtering algorithm to help users of online reputation systems avoid the misleading of dishonest ratings. This algorithm evaluates the trustworthiness of ratings by comparing the raters’ opinions with the opinions of the evaluator, and gives the ratings proper weights before including them into the final judgment. Different weighting functions are applied to positive and negative ratings adaptively so that the weights can better capture the characteristics of various types of malicious raters. Simulations prove that the proposed algorithm can effectively avoid misleading ratings, minimize their bad influences on trust evaluation, and help users make more reliable trust decisions from a personal point of view.