MRHN: Hypergraph Convolutional Network for Web API Recommendation

Abstract

With the development of service-oriented computing, Mashup technology has emerged that uses web API as reusable components to create new products. How to achieve efficient and accurate service recommendation has attracted the attention of researchers in the field of service computing. The call relationship between mashups and APIs in real service data is intricate, and the information carried by the service further increases the complexity of the relationship between them. Most existing mashup recommendation models hardly mine such complex relationships effectively. To this end, this paper proposes the MRHN method. This method uses motifs to extract the hypergraph structure from services. While studying the complex relationship between service data, it also solves the problem of data sparsity, and uses the hypergraph convolutional network to extract the features of Mashup. Further, the weights of channels are adjusted using channel error attention mechanism. Finally, the performance of the proposed method is comprehensively evaluated. The experimental results show that compared with the existing service recommendation methods, the proposed method has significantly improved in terms of evaluation indicators such as NDCG and HR.