Research On Web Service Recommendation Technology Based On Knowledge Graph

Web service recommendation as an effective solution to the problem of service information overload has become a popular direction in the field of service computing.Most of the current service recommendation research is based on collaborative filtering algorithms The data sparse problem,which has a large negative impact on the calculation of the similarity of services or users,and cold start problem faced by this kind of algorithm reduce the accuracy and the diversity of the recommendation.In this paper,the Web service recommendation technology is deeply studied from the aspects of Web service knowledge graph construction,vectorization algorithm of entities and relationships,and service recommendation algorithm.The purpose is to use the rich correlation between entities contained in the entity vector representation to improve the accuracy and diversity of recommendations by solving data sparseness and cold start problems in some extent.Based on the analysis and research of Web service recommendation,knowledge graph,entity vectorization and entity similarity measurement and other theories and technologies,this paper makes a deep study on Web service recommendation technology based on knowledge graph.First,a Web service recommendation framework is designed.This framework gives the entire process of Web service recommendation based on knowledge graph,which achieves the recommendation of relatively accurate and diverse Web services for users by utilizing the rich association relationship between users and services and user classification.the goal of.At the same time,the components of the Web service knowledge graph are analyzed and the Web service knowledge graph is constructed.The entities and relationships in the Web service knowledge graph are identified and extracted through real Web service related data.The Neo4j graph database is used to complete the storage and visualization of the Web service knowledge graph.Then,an improved entity vectorization algorithm,imp-TransH,is proposed and compared with other related algorithms.Because there are lots of many-to-many relationships between entities in Web services,this algorithm optimizes the negative sample collection algorithm of the translation model algorithm TransH to express entities and relationships as numeric vectors but retaining their semantic information.Finally,a Web service recommendation algorithm based on knowledge graph,WS-KG,is proposed and experiments are conducted to verify the effectiveness of the algorithm.It applies the imp-TransH algorithm to generate entity and relation vectors,calculates the similarity between users and services according to these vectors,predicts users'ratings of unfocused services based on user nearest neighbors and service nearest neighbors,classifies users into categories such as positive or negative according to their interest distribution,and generates a recommendation list for the user based on servies' ratings and the user's category.Experimental results show that the recommentaion accuracy and diversity of the proposed WS-KG algorithm in this paper is higher than compared algorithms.This paper uses the rich service semantic information in the knowledge graph to make up for the lack of semantic information of services and users in the service recommendation algorithms based on collaborative filtering,the data sparse problem and the cold start problem can be solved to a certain extent.