Research On Web Service Discovery Mechanism Based On Deep Learning And Behavior Path Graph

With the rapid development of the Internet and information technology,the base of netizens has become larger and larger,and the Internet penetration rate has grown rapidly,which has simultaneously expanded basic services.As more and more services are registered on the Internet,users are faced with a large number of Web services with similar functions and similar types.At the same time,the influx of services with different levels of quality has created service overload and chaos.It is difficult for users to find Web services that meet their own preferences,and it is also a waste of resources for service providers.The advent of Web service discovery methods has greatly alleviated such problems.Existing service discovery methods can be roughly divided into two categories:s syntactic discovery and semantic discovery.Syntactic discovery methods mainly focus on grammatical keyword matching,but there are problems such as ambiguity of keywords and insufficient keyword provision,which limit and reduce the accuracy of service discovery.Semantic discovery methods try to overcome these problems by combining the power of ontologies,that is,providing semantics through ontologies to obtain formal service descriptions.However,with the continuous growth of the number of web services,there are more and more web services with the same function,which makes it extremely difficult for the semantic discovery method combined with ontology matching to capture accurate user functional requirements,resulting in service discovery.accuracy is reduced.To improve the accuracy of service discovery,researchers use text matching for service discovery.This paper has conducted a lot of research on the existing service discovery methods based on this matching.It is found that in the existing deep learning methods based on text matching,for the feature extraction process,most methods are developed based on a single dimension such as word frequency,and these methods are insufficient.to accurately capture the comprehensive characteristics that represent the service.Also,the neural network model used to convert service matching to text matching only takes text representations as input and provides too little lexical interaction information.Aiming at the above problems in text matching,this paper proposes a service discovery method based on multi-dimensional representation neural network.The main research contents are as follows:(1)Aiming at the defects of insufficient feature extraction in existing text matching,this paper develops a service discovery method based on multi-dimensional representation neural network.First,feature extraction is performed from multiple dimensions during data processing to make up for the lack of single-dimensional feature capture.For each keyword,the corresponding TF-IDF,Word2Vec and ELMo representations are generated to capture word frequency,static contextual features and dynamic features.contextual features.In addition,the cosine similarity is calculated based on the word pairs between the query and the web service and a multi-dimensional similarity matrix is constructed,which is used as the input of the CNN.Finally,in order to verify the proposed method,this paper designs three sets of comparative experiments from the perspectives of accuracy index and error index.The method is low,and the overall experimental results show that the method proposed in this paper has advantages in service discovery performance and accuracy.(2)In order to make up for the defect that text matching cannot be used in special scenarios,this paper draws on the idea of service process and maps the service process into a service process model diagram.With the help of graph theory,the service process model graph is decomposed into behavior path subgraphs,and the kernel function is used to calculate the similarity of the subgraphs,so as to transform the service matching problem into the behavior path subgraph similarity problem to solve.After comparative experiments,it is proved that the recall rate and precision rate of the method adopted in this paper reach 92.1%and 90.5%,respectively,which are higher than the baseline VFGCN method and clustering bipartite graph method.(3)Finally,this paper designs and implements a prototype system of service publishing and discovery based on B/S architecture based on the model method proposed in chapters 3 and 4.The system is written in Java language,SpringBoot is used as the back-end framework,LayUI is used as the front-end framework,MySql database is used for the data layer for persistence,and Redis is used to cache hot data.