Hybrid high-order semantic graph representation learning for recommendations

Abstract The amount of Internet data is increasing day by with the rapid development information technology. To process massive amounts and solve overload, researchers proposed recommender systems. Traditional recommendation methods are mainly based on collaborative filtering algorithms, which have sparsity problems. At present, most model-based algorithms can only capture first-order semantic cannot high-order information. above issues, in this paper, we propose a hybrid graph neural network model heterogeneous graphs extraction, models users items respectively learning low-dimensional representations for them. We introduced an attention mechanism to allow understand corresponding edge weights adaptively. Simultaneously, also integrates social learn more abundant performed some experiments related datasets. Our method achieved better results than current advanced models, verified model’s effectiveness.