Research On Web API Recommendation Algorithm Based On Knowledge Graph Neighbor Information Propagation

The rapid growth of the number of Web APIs not only brings sufficient resources to developers,but also increases the difficulty of decision-making.Web API recommendation algorithm came into being and has become an important and meaningful research direction.The inherent problem of data sparsity seriously affects the accuracy of recommendation algorithm.Previous studies have proved that auxiliary information can alleviate data sparsity.This paper uses knowledge graph as supplementary information to enhance Web API recommendation,Web API recommendation is studied from the aspects of knowledge graph neighbor information propagation method and recommendation algorithm.The main work is as follows.Firstly,research on knowledge graph neighbor information propagation method,which is used for enhancing Web API recommendation.Aiming at the natural data sparse problem in Web API recommendation that cannot be effectively solved by existing methods based on collaborative filtering and deep learning,consider using the knowledge graph neighbor information propagation method to integrate knowledge graph.The overall framework of Web API recommendation based on knowledge graph is described.Based on the analysis of the web API ecology,the top-level entity relationship model of the Web API knowledge graph is designed,and use the real Web API related data to extract entities and relationships to complete the construction of the Web API knowledge graph.Find the users and Web API entities corresponding to the historical interactive data in the Web API knowledge graph and take them as the starting point.Automatically obtain the neighbor information of the multi hop triple structure through the relationship,capture the hidden structure information and deep semantics of the knowledge graph,and make up for the defect of sparse historical interactive data.Secondly,research on knowledge graph enhanced Web API recommendation via neighbor information propagation method.The existing Web API recommendation based on deep learning ignores the rich structural and relational semantic information in the Web API ecosystem.In addition,the mining of effective features is highly dependent on artificial feature engineering,which limits the model scalability and prediction accuracy Therefore,we propose a knowledge graph enhanced Web API recommendation model via neighbor information propagation,KGWARec.This algorithm uses neighbor information propagation method,and is based on the historical interests and characteristics of users contained in the historical interaction data between users and the Web APIs,combined with the multi-layer neighbor information of the knowledge graph,iteratively calculate the representation of the user and the Web API which after the propagation of each layer,and then,the representation of each layer is superimposed to obtain a multivariate representation of the user and the Web API integrating multiple types of information,which is used to calculate the prediction probability of the user using the targeted Web API.Finally,experiments are carried out to prove that the recommendation performance of KGWARec in the case of sparse data,and the accuracy,stability and effectiveness of KGWARec method are analyzed.