Xie Jin,∗,∗∗ Xianfeng Weng,∗∗∗ Xinru Fan,∗∗∗∗ Mohamad Fadli Zolkipli,∗∗ Yufeng Chen,∗ Zhengtao Xiang,∗ and Minghao Yue∗
Commercial robots, graph neural network (GNN), attention mechanism, recommendation algorithm
In the commercial field, commercial robots can recommend high- quality products to customers. Aiming at the limitations of cold start and sparsity in current recommendation models, a graph neural network recommendation method combining entity similarity and relationship aggregation (ERF-GNN) is proposed. The neighbour set is generated by using similarity, and the information interaction on the knowledge graph is recorded for expansion to enhance the user feature representation. At the same time, the attention mechanism is used to integrate the neighbourhood weight into the entity, enhance the embedding representation of the node, and improve the recommendation accuracy and richness of the robot. The experiment shows that this model has better recommendation performance compared to other benchmark models on the Book-Crossing and Yelp.
Important Links:
Go Back
