Graph Context Target Attention Graph Neural Network for Session-based Recommendation

Abstract

Session-based recommendation is nowadays increasingly popular in e-commerce, aiming at predicting the next action of a user under anonymous sessions. Previous research methods on session recommendation model the temporal information inherent in a session as a sequence or graph, however, they disregard the session's graph context information, as well as the relationship between the user and the target object, which affects the accuracy of the recommendation. To obtain the rich graph context information in session recommendation and the intrinsic connection between target items and users, we propose a graph context target attention graph neural network for session-based recommendation, which uses a self-attentive network and graph neural network to extract the item embedding of graph context information; the target attention then adaptively stimulates various user interests. Experimental results on two real-world datasets demonstrate that our proposed model outperforms other comparison algorithms on the evaluation metrics of Recall@20 and MRR@20 in session-based recommendation.