Research On Matching And Evaluation Technologies For WEB Service Discovery

With the continuously improvement of industrial standards and the maturity of Enterprises Application Platform of Web Service,it is a new trend to corporate and integrate the various applications as the form of Web Service.Therefore,how to achieve the web services concerned with the users' requirements from the huge amount of Web Service becomes a main task of the research field of Web Service Discovery.Current technologies of Web Service Discovery include the service matching and service evaluation.The service matching uses is to measure the difference between the users' functional requirement and the function provided by a web service,and achieves the reliable web services.The service evaluation is to measure the online performance of Web Service from the results of the web services matching,and achieves the available web services.The service matching and evaluation are interrelated and interact with each other.According to the problems of current technologies of the service matching and evaluation,the detail works are listed as below:(1)In the matching process when a single service satisfying with the users' requirements,there are some problems including the difficulties of ontology extraction,lack of field feature for the similarity calculation between the users' requirements and the services ontologies,and the low accuracy of matching results from the matching algorithm.To solve the problems,a single service matching method based on the ontology is proposed.Firstly,the method defines the ontology models and formalizes the proposition of the single service matching.Secondly,with the restriction of proposition,the matching framework is designed including the processes of the ontology extraction,ontology similarity calculation and matching algorithm.Thirdly,the detail algorithms of the ontology extraction,similarity calculation with the field features and achieving optimal matching solution module are introduced.(2)As the single service matching is difficult to effectively utilize the services in the services library to cooperatively implement the complex business requirements,and high time complexity of the current matching algorithms along with the concept of the compound service matching,a compound service matching method based on coevolution algorithm is proposed.Firstly,the method formalizes the proposition of the compound service matching in the light of the proposition of the single service matching,and transforms the original matching problem into the intelligent problem by the proposition of compound service matching.Secondly,the equivalent transformation procedure of matching solution is designed to convert the matching solution into the form of the evolution basis that could be processed by Intelligent Algorithm.Thirdly,by the idea of coevolution and the advantages of Particle Swarm Optimization and Simulate Annealing,the flow of coevolution algorithm is proposed to solve the compound service matching problem within a limit time.(3)Current subjective service evaluation technologies for the target services based on KNN is difficult to quantify of the time utility of the subjective historical feedback,and hard to set the width of evaluation windows and control the malicious feedback.To solve these problems,a web service evaluation method based on variant time utility is proposed.Firstly,the method uses System Dynamics to build the complex quantification method of the time utility,as to achieve the variant time utilities for different target services.Secondly,from the results of the quantification method,the suitable width of the evaluation window of KNN is adaptively calculated.Thirdly,the statistics of subjective feedbacks of user is used to estimate the relative confidence interval of normal feedbacks,which filter and improve the malicious feedbacks and reduce the side-effects on the prediction accuracy.(4)Current objective service evaluation technologies for the target users evaluation prediction based on collaborative recommendation is hard to quantify the historical invoking information,ignore the tightness relations between the target users and different services,and could not predict when there is insufficient historical invoking information.To solve these problems,a web service evaluation method based on the time utility recommendation is proposed.Firstly,the quantification of the time utility of the historical invoking information is designed on the basis of the current quantification approaches.Secondly,under constrains of the time utility,the tightness between the target users and different services are measured and the tightness is treated as weights to optimize the similar users' calculation to the target users.Thirdly,in the light of hypothesis that nearby users shares the similar link environments,the process to find neighbors to the target users based on the geographical information are proposed,which improves the accuracy of evaluation prediction.