Cross-platform Recommendation Considering Common Users’ Preferences Based on Preference Propagation GraphNet

In this paper, we propose a cross-platform recommendation (CPR) considering common users’ preferences. Conventional CPR approaches focus on a situation where users utilize only one platform. However, with the growth of subscription services in real-world applications, more users utilize multiple platforms in search of platform-specific contents. Therefore, we construct a CPR method assuming that users utilize multiple platforms. Inspired by the concept of the cross-domain recommendation that leverages other domains’ information via common users, the proposed method enables propagation of information from another platform through interactions of common users. As a result, the proposed method can estimate users’ preferences and improve the recommendation performance even if they have a few interactions. Experiments on an open dataset show the effectiveness of the proposed method.