Research On Program Mining Based On Component Resources Warehouse

Program Mining is a new model of network computing in network environment which implements the on-demand service.Its basic idea is to analyze users' requests, retrieve required components and assemble them into an integral program to meet users' needs and achieve service creation and customization dynamically.According to the order of dealing with users' requests,the process of program mining is divided into five steps:submit users' request;analyze users' request and dissect functions;search and retrieve components;tailor and assemble components; verify and deploy program.Among them,retrieving components is a key one.On the basis of the Internet and local component resource warehouse,the paper researched mainly the framework model of program mining system and the component retrieval strategy.In the multi-Agent platform,the framework of program mining system was designed according to importing the method of case-based reasoning(CBR) and domain ontology and user interest.Meanwhile,considering the problems of intelligence and personality in web service,a new retrieval mechanism based on domain ontology and user interest was present,which combined domain ontology and user interest on the basis of the search mechanism of facet and keywords to retrieve required components.Except that,the program mining system was implemented,and it proved feasibility of system and effectiveness of the component retrieval method with an instance.Specific studies are as follows:CBR was applied to the program mining system;the system framework was improved through the establishment of user interest model and domain ontology repository.The system has the characteristics of intelligence and personality of active service.Ontology has features of semantic and cohesion,so ontology was used to describe components to define ontology feature domain model.The acquiring method of user interest information was put forward.On the basis of these,a retrieving method based on domain ontology and user interest and an algorithm of searching associated component were present. Finally,the Mining System was developed on the platform of Microsft.Net, which included a number of subsystems,taking into account the complexity of the procedure of mining system,the paper detailed the subsystem of retrieving components,and an instance with components in E-Commerce field proved feasibility and validity comparing with the retrieval methods based on keywords and facet.The experiments show that the method has higher precision and recall than others for large-scale components library.