Graph Neural Networks with Category Embedding for Session-based Recommendation

Session-based recommendation plays an important role in e-commerce to recommend items of interest for users. However, existing session-based recommendation algorithms only use the item ids in the data without using other information, which makes their recommendations less effective. To solve the problem, this paper proposes a Graph neural networks with category embedding for session-based recommendation. First, the model embedded the item category into the item id to get the item representation. Then the graph neural network is used to learn the transfer relationship between items, and the final representation of the current session is obtained through the attention mechanism, which is used to complete the prediction, solving the above problem. Extensive experimental results on two public datasets, UserBehavior and Tmall, show that the model is better than existing methods.