A multigranularity deep recommendation model based on graph matching

Graph neural networks can effectively capture data features of higher-order topologies. And it is widely used in deep recommendation models. Existing recommendation models are prone to the problem of inaccurate recommendation sequences caused by user interest drift. In order to optimize the problem effectively, this paper proposes a multigranularity deep recommendation model based on graph matching (MG-GMCF). First, the data samples are divided by attributes, and then, a multi-granularity feature extraction structure is constructed. After interacting through the topology of different structures, the extracted features are fused at multiple granularities to obtain the feature graph representation. Finally, graph matching is then utilized to predict the recommendation results. Experiments are conducted on three public datasets to verify the effectiveness of the proposed model.