Multiobjective and Multistakeholder Recommender Systems

Abstract

As an effective information extraction tool, recommender systems(RSs) can effectively provide users with content strategies from a large amount of data. The traditional RS can discover the unknown products of users and satisfy their tastes. However, the preferences of other RS participants should also be considered, such as the platforms. The platform’s objective is different from that of users, and they want to maximize profits. In this paper, we adopt a multiobjective model MSMO for multistakeholders, in which customer relevance and profit of the platform are taken into consideration. By applying four evolution techniques, we are able to find Pareto front as optimal solutions. The solutions can keep the balance among multiple stakeholders. Experiments on a real-world data set reveal that our proposed model can significantly promote profit with little sacrifice in individual preference.