MKGPC: Multimodal Knowledge Graph Propagation for Recommendation Systems

With the exponential increase of information available on the Internet, the problem of information overload has become more and more urgent. Recommendation systems have been proposed to solve this problem by recommending a small subset of information from a large pool of data for users to quickly digest. Collaborative filtering has been the main direction of recommendation algorithm research, but it faces challenges such as the cold start problem and sparse interaction data. To address these challenges, researchers have introduced auxiliary information such as point-of-interest, comment, social networks, and contextual information. Knowledge graphs have been introduced as auxiliary information to improve the performance of recommendation systems. Knowledge graphs can fuse multiple data sources, expand the semantic information of data, and infer more potential information of data. However, current knowledge graph-based recommendation systems tend to ignore multimodal information. In this paper, we propose a new model, termed as MKGPC, that utilizes the propagation of information on the multimodal knowledge graph and considers multimodal information fusion. We compare the proposed model with state-of-the-art models. The proposed model shows promising results on two real-world datasets.