Bundle Service Recommendation Based on Graph Representation Learning

Service recommendation aims to quickly select appropriate services from a large number of Web services to meet user needs. However, existing service recommendation methods mostly focus on improving the accuracy and relevance of recommendation results, ignoring the diversity of recommendation results, thus leading to the singleness of recommended content. Therefore, in this paper, we propose a bundle service recommendation method based on graph representation learning (BSR-GRL). We first construct a heterogeneous information network centered on user-service graph, user-bundle graph and hypergraph, and divide them into single-service view and bundleservice view. Then, the graph convolutional neural network is utilized to perform representation learning on different views, and on this basis, BSR-GRL exploits the InfoNCE loss function for cross-view contrastive learning. Finally, we combine hypergraph embedding vectors, comparative learning results and semantic coding for diverse service recommendation. The experimental results show that the proposed BSR-GRL model is efficient. In terms of ILR and H, the values improve by 3.87% and 2.32% over the optimal baseline, respectively.