Graph matrix completion for power product recommendation

The recommendation system has achieved great success for decades, some are still not solved well, especially in data sparseness. Graph convolution network(GCN) can be well applied to recommender systems, but the previous methods only consider vector representations of users and items. The latent in user-item interactions is not encoded in the embedding process. As such, the resultant embeddings may not be sufficient to capture the collaborative filtering effect. we propose a new Graph matrix completion(GMC) framework, which add user-item interaction information and side information to the graph auto encoder network, and capture differentiable information transmission on bipartite interaction graph. Multi-layer decoder is adopted to optimize loss for better map the potential representation of user/item. Our model is evaluated on multiple benchmark datasets and outperforms the state-of-the-art algorithm.