Grid Application Scheme Generation System Based On Case-based Reasoning

With its powerful resources sharing capability and coordinating question solvingability, grid technology has attracted many people’s attention. Applying grid technology todaily life has become a common vision. However, the dynamic, heterogeneous, distributedcharacter of grid resources and the complex, flexible, changeful character of gridtechnology make the application of grid technology become a difficult work. In order tosolve this problem, this thesis constructs a scheme generation system by the case-basedreasoning technology. Through knowledge reasoning, the system will generate a gridapplication scheme which is consistent with the industry standards and meets therequirement of users, what will reduce the difficulty of grid application.Grid technology involves much knowledge and has a very high demand of systemoverall performance, while the traditional rule-based reasoning technology has difficultyin the gaining of knowledge and rules, and the overall performance of it is also poor.Therefore, we will use the case-based reasoning technology to construct the gridapplication scheme generation system in this thesis, in order to avoid the problem above.In this thesis, we first analyze the principle of IDSS and case reasoning, design amodel of grid application scheme generation system based on the analysis above, andanalyze the principle of it. Then, we make an in-depth study on the reasoning mechanismof grid application cases.CBR involves case expression, problem description, case retrieval, case reuse andcase study. This thesis focuses on the case expression and case retrieval of the gridapplication cases.About the case expression, the thesis proposes the grid application case generalarchitecture as the standard of case expression based on the analysis of grid architectureand grid application type, which will ensure the standardization, structuration and generalization of system cases.About the case retrieval, according to the characteristic attributes of the gridapplication cases, the thesis designs a case retrieval model of grid application cases basedon the nearest neighbor strategy, proposes a new case similarity measure methodcombined with the nearest neighbor measure model and fuzzy analytic hierarchy process,which improves the accuracy of similarity measure.Based on the study above, we design and implement the grid application schemegeneration system based on CBR, then elaborate the work flow of the system and analyzethe performance of the system. Test results indicate that the system has the ability togenerate the grid application scheme which is consistent with the industry standards andmeets the requirement of users, achieves the expected target.