Knowledge Graph-based Intent Network-Enhanced Web Services Recommendation

APIs recommendation for Mashup creation is becoming a hot topic in service computing. However, most existing APIs recommendation approaches fail to model the relationship between Mashup and API at a fine-grained level of intent, much less exploit relational dependencies to preserve the semantics of remote connections. Therefore, we propose an API service recommendation method based on the interactive learning intent of knowledge graphs for Mashup creation in this paper. This approach first creates a knowledge graph of APIs using natural language processing. Then, it uses the interaction learning intent model in the knowledge graph to model the intent of Mashups and APIs, obtain more independent Mashups and API intents by filtering intent independence and distance correlation, and recursively integrate them to gain intent-aware Mashup and API representations. Finally, the obtained Mashup and API representations are subjected to inner product operations to predict and generate API recommendation results. The experiments on real dataset collected from Programmable Web are performed, and the experimental results show that the recall of the proposed method have 11.35%, 15.72%, 21.33% and 9.36% improvement, the NDCG of the proposed method have 9.36%, 22.13%, 5.25% and 31.26% improvement, respectively, compared to Deep-FM, KGAT, CKE and R-GCN.