In the age of information explosion, Internet facilitates product searching and collecting much more convenient for users. However, it is time-consuming and exhausting for users to deal with large amounts of product information. In response, various recommendation approaches have been developed to recommend products that match users’ preferences and requirements. In addition to the well-known collaborative filtering recommendation approach, the trust-based recommendation approach is the emerging one. The reason is that most of online communities allow users to express their trust on other users. Based on the analysis of trust relationships, the trust-based recommendation approach finds out and consults the opinions of more reliable users and therefore makes better recommendations. Existing trust-based recommendation techniques consider all trust relationships in a given trust network equally important and give them the same trust strength. However, in a real-world setting, trust relationships may be of various strengths. In response, in this study, we propose a mechanism for trust strength estimation on the basis of the machine learning approach and estimate the trust strength for each existing trust relationship in a given trust network. To overcome the sparsity of the trust network, we also develop a modified trust propagation method to expand the original trust network. Finally, we perform a series of experiments to demonstrate the performance of our trust-based recommendation approach based on the trust strength estimation mechanism. Our empirical evaluation results show that our proposed approach outperforms our benchmark techniques, i.e., the traditional collaborative filtering approach and the original trust-based one.
Recommendation Systems, Collaborative Filtering, Trust Network, Trust Relationship, Trust Strength, Machine Learning
ISBN: [978-1-86435-644-1]; Full paper
Wei, Chih-Ping; Chen, Hung-Chen; and Liang, Ming-Kai, "Estimating Trust Strength For Supporting Effective Recommendation Services" (2011). PACIS 2011 Proceedings. 212.