Leveraging Social Trust Matrix Factorization for Enhanced Recommender Systems

Abstract

The growing of digital industry such as e-commerce has raised the issue of users facing the challenge in determining
their needs due to the increasing number of available products options. We believe that the user’s network or user’s neighborhood can help to alleviate this issue by recommending items that the closest neighbors have purchased, yet we argue that the indirect influence of trust should also be considered. In this study, we adopt an approach that uses trust matrix factorization
and collaborative filtering models to identify the influence of trust amongst users. This trust data will be mined to identify pattern in which a target user is influenced by other users that are in her networks. Trust matrix factorization is used to model the level of trust between users based on existing social relationships. Meanwhile, the collaborative filtering model is used to identify patterns of similarity in consumer characteristics. Through this approach, we aim to see which social recommendations are better suited to represent user preferences. Here, we proved that the proposed model offers more personalized and relevant recommendations by considering the indirect influence of trust on top of trust matrix factorization and collaborative filtering models. The results of this study have important implications for the development of effective and accurate recommendation systems. By highlighting the role of trust influence, this research provides valuable insights into understanding user interactions and designing better recommendation systems in the future.