TCFACO: Trust-aware collaborative filtering method based on ant colony optimization. (15th March 2019)

• Record Type: Journal Article
• Authors: Parvin, Hashem; Moradi, Parham; Esmaeili, Shahrokh
• Is Part Of: Expert systems with applications. Volume 118(2019)
• Page Start: 152
• Page End: 168
• Publication Date: 2019-03-15
• Subjects: Collaborative filtering -- Recommender systems -- Social trustinformation -- Similarity measures -- Ant colonyoptimization; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2018.09.045
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: A social recommender system method called TCFACO is proposed. Trust statements are used as a side information to deal with the data sparsity and cold-start issues. TCFACO uses available rating values along with social trust relationships to rank users. TCFACO uses ACO to choose a set of valuable users with their associated weights. Final selected users are used in the rating prediction considering their similarities to the target user. Abstract: Recommender systems (RSs) aim to help users to find relevant information based on their preferences instead of searching through extensive volume of information using search engines. Accurate prediction of unknown ratings is one of the key challenges in the analysis of RSs. Collaborative Filtering (CF) is a well-known recommendation method that estimates missing ratings by employing a set of similar users to the target user. An outstanding topic in CF is picking out an appropriate set of users and using them in the rating prediction process. In this paper, a novel CF method is proposed to predict missing ratings accurately. The proposed method called TCFACO uses trust statements as a rich side information with Ant Colony Optimization (ACO) method. TCFACO consists of three main steps. In the first step, users are ranked considering available rating values and social trust relationships. Then, in the second step, the ACO method is utilized to assign proper weight values to users to show how they are similar to the target user. A set of top similar users is filter out in the third step to be used in predicting unknown ratings for the target user. In other words, to speed up identifying similar users, the proposed method first filters out a majority part of dissimilar users and then runs the ACO on only a reduced set of users to weight them. Several experiments were performed on three real-world datasets to evaluate the effectiveness of the proposed method and the results show that the proposed method performs better than the state-of-the-art methods.