Calculating trust by considering user similarity and social trust for recommendation systems

Abstract

This research presents a trust assigning method for recommendation systems by considering a user similarity and social trust. Herein, the proposed method consists of three main processes, namely trust calculation, neighbor filtering, and items rating prediction. To evaluate, the FilmTrust dataset was used to verify its prediction performance. The results shown that the significant measures, such as the mean absolute error (MAE) and percentage of accuracy, they were around 0.197 and 80% with a trust walk in a social network, λ = 5, respectively.