Research And Implementation Of Web Service Matching Model Based On Semantic

In the upsurge of Internet technology and its application, Web services, as a newWeb application model， has becoming an ideal resource access standard in theintegration and sharing of applications and resources which belong to heterogeneousplatforms. However, with the growing number of Web services, how to find servicesthat meet users’ needs in a flood of Web services is becoming a widespread concern.Semantic level matching can solve the common problems such as low recall rate andlow precision rate which are inevitable in syntax level matching based on keywords,therefore, semantic matching has been recognized by the industry. However, there issome one-sidedness in the current semantic-based Web service matching methods. Forinstance, lack of consideration of context information and the uncertainty of serviceproperties. In addition, the greedy algorithm which is applied in common matchingalgorithms will lead to False Positive and False Negative. This thesis studies the therelated semantic Web service matching model and focuses on how to describe Webservice and service request, how to match service request and service advertisment,and how to recommend service meet users’ demands mostly to the user.This thesis proposed a hybrid Web service matchmaking algorithm on the basis ofthe existing methods. The proposed approach aims to combine the merits of fuzzyrough set theory and merits of bipartite graph matching. First, function elements, suchas input properties and output properties are extracted from Services Profile of serviceadvertisements and service requests, respectively. In order to address uncertainty inservice properties, the fuzzy rough set theory is adopted to eliminate redundantfunction properties. In addition，according to the remaining properties, a bipartite graphis constructed. Hence the calculation of similarity degree will be switched to theproblem of extended optimal matching for bipartite graph. Function information suchas input, output, precondition and effect and non-function information such as userpreference, Qos and context are considered in the presented method. Thus the recallrate and accuracy are improved.Finally, on the basis of the hybrid method, this thesis designed and realized a Web service matchmaking model, and also verified the matchmaking model. Throughanalyzing and comparing the experimental results with other algorithms. It shows thatthe proposed hybrid method can improve the recall rate and context information caninprove the accuracy of the matching results.