Graph Neural Network Session Recommendation Assisted with Contrastive Learning
Session based recommendation(SBR)is a challenging task aimed at recommending items based on anonymous behavior se-quences.This article proposes a new method called Graph Neural Network Session Recommendation Assisted with Contrastive Learning(CLGNN).Based on the graph attention mechanism,contrastive learning is used to assist training in order to obtain better recommended results.Specifically,CLGNN uses attention mechanism to learn item embedding on the session graph,and then aggregates items within the session to generate session embedding.Finally,generate recommendations using session embedding and candidate item embedding to calculate scores.At the same time,contrastive learning is used to optimize item embedding space.Based on several common evaluation index methods,experiments were conducted on two real datasets,and the results showed that the model recommendation performance in this article is good.
NETL
NSTL
万方数据
