17 / 2023-08-21 10:26:45

Fusing Item Knowledge And Sequential Graph Information For Session-based Recommendation

session-based recommendation,graph neural network,knowledge graph,memory network,deep learning

Special Sessions > SS02 - Intelligent Robot, Artificial Intelligence Methods and Its Applications

As one of the application fields of artificial intelligence, session-based recommendation aims to provide users with personalized suggestions and recommendations to meet their diverse short-term needs. However, traditional session-based recommendation faces a series of challenges in capturing dynamic preferences in complex sequences and providing interpretability. To this end, this research proposes the KGSrec session-based recommendation model, which combines item knowledge and sequence diagram information. The gated graph neural network is utilized to model session sequence features, while the TransR knowledge representation method is used to model item entity and attribute features. Furthermore, the two types of features are integrated via the key-value memory network approach, and the session preference matrix is applied to give users with attribute-level recommendation explanations. The experimental results on two public da tasets of LFM and Movielens show that, when compared to the classical model, the Recall@10 metric of KGSRec on the two datasets is increased by 6.3% and 2.5%, respectively, and the NDCG@10 metric is increased by 0.46% and 0.36%, demonstrating the model's effectiveness.