Research On Case-based Reasoning In Decision Support System

Case-Based Reasoning(CBR) is an important method in solving problem in the area of artificial intelligence, it shows great vitality because of it's special reasoning style and success application, more and more people show wide range of interest to CBR. CBR can be understood as a method of using the past cases and experience to solve new problems, it also can be considered as an analogy inference from one case to another one.With the development of computer science, using the computer for decision support is becoming a trend. In the view of designing the module of CBR in decision support system(DSS), this dissertation researched the key technology of CBR such as case representation, casebase, case retrieval.The main contents are in brief as following:(1)On the basis of the introduction to the origins, development and research status of CBR, the basic method, system framework and the key technology of CBR are discussed in detail in this dissertation.(2)According to the problem of designing the complicated temporal cases and casebase in some specific areas. Based on using framework as the basic method for representing case, we described the case with the attribute of time; It can reduce the case base's size in the condition of maintain the crucial information of cases by using correction analysis to select feature attribute and principal components analysis to rebuild feature attribute.(3) For the purpose of raising the efficiency of case retrieve, firstly induction index method is used for reducing the searching space, and then nearest neighbor method is used for searching the most similar case. In order to solving the problem of setting the feature weight when case is retrieved, a combined method for setting feature weight based on entropy was proposed in this dissertation. The method's validity is testified by experiment.(4) The implementation of CBR module in disaster rescue decision support system was studied finally.