ECAT: an Enhanced Confidence-aware Trust-based recommendation system

Abstract

One of the most widely employed recommendation approaches is collaborative filtering. There are some limitations to this approach, such as the cold-start user problem. Trust-based recommendation approaches are solutions to prevent traditional collaborative filtering-based approaches generating a poor recommendation. We have introduced a unique recommendation approach in our conference paper, called Confidence-aware Trust (CAT), which considers a confidence estimate in both direct and indirect trust calculations. However, the CAT recommendation approach does not involve some vital aspects, namely: the number of neighbors of active users/ target items, consistency in rating values, and the level of confidence in the predicted rating value as well. To address these aspects, we further propose an innovative approach, called an Enhanced Confidence-aware trust (ECAT), which improves the CAT recommendation approach. The Movielens dataset was evaluated. The experimental results show an improved performance by ECAT over its counterparts with respect to the accuracy of recommendations, especially tackle the issue of cold -start users.