A Collaborative Filtering Recommender Approach by Investigating Interactions of Interest and Trust

Collaborative filtering-based recommenders operate on the assumption that similar users share similar tastes; however, due to data sparsity of the input ratings matrix, traditional collaborative filtering methods suffer from low accuracy because of the difficulty in finding similar users and the lack of knowledge about the preference of new users. This paper proposes a recommender system based on interest and trust to provide an enhanced recommendations quality. The proposed method incorporates trust derived from both explicit and implicit feedback data to solve the problem of data sparsity. New users can highly benefit from aggregated trust and interest in the form of reputation and popularity of a user as a recommender. The performance is evaluated using two datasets of different sparsity levels, viz. Jester dataset and MovieLens dataset, and are compared with traditional collaborative filtering-based approaches for generating recommendations.