Research of collaborative filtering recommendation algorithm based on trust propagation model

Abstract

Traditional collaborative filtering recommendation algorithm is one of the methods to solve the information overloading problem in E-Commerce. However, there are four urgent problems in this algorithm namely data sparse, cold start, attack-resistant and scalability. This paper makes a trust propagation model called TPM; proposes a hybrid index called TS index and a novel collaborative filtering recommendation algorithm called TPCF using TPM and TS index. The results of experiments using the dataset of Epinions.com, a popular ecommerce review website, show that TPCF is more attack-resistant and improves the precision and coverage rate compared with the traditional collaborative filtering recommendation algorithm using Pearson's correlation coefficient. TPCF has a better performance against the traditional collaborative filtering recommendation algorithm on the problems of data sparse, cold start and attack-resistant.