Cbr-rbr Fusion Model Construction For Medical Application

China's medical resources are insufficient. The imbalances and differences of medical diagnosis and treatment among regions and medical institutions make many patients unable to access a high level of standard healthcare and medical quality are difficult to be effectively guaranteed. Particularly, the diagnosis and treatment of such a high risk disease, heart disease, need to be improved. To address the shortage of medical resources and poor quality of health care, medical decision support system based on CBR or RBR provides an effective way. However, CBR and RBR have their advantages and disadvantages, and there are few medical decision support systems that can effectively integrate the two and are applicable to heart disease diagnosis.Based on analysis of research results related to theoretical methods of case feture selection, CBR case retrieval, RBR reasoning and integration of CBR and RBR, we construct a CBR-RBR Fusion Model for medical application considering the charac-teristics of medical diagnosis. Firstly, rough set theory is used for feature selection in this model. Secondly, optimal weights for similarity calculation of the cases are cal-culated by integrating the value of information gain and Rough Sets of medical record attributes. Thirdly, we propose a new algorithm used to calculate remote nearest dis-tance of K-D tree, which improves K-D tree search efficiency. And the improved K-D tree algorithm is used for CBR case retrieval. Fourthly, we put these cases whose si-milarity of CBR case retrieval is below the set threshold into RBR reasoning, where the Bagging algorithm and C4.5 decision tree are used to train and integrate the cases, that is building a Bagging-C4.5 integrated model of multiple classifiers. As a result, the performance of RBR reasoning is effectively increased. Finally, this model is ap-plied to heart disease problem. We use the UCI-Cleveland heart disease database for training and predicting. And the comparison between this model and other related re-search results indicates that the proposed model has high accuracy, sensitivity and specificity in heart disease diagnosis and that the proposed model has a good prefer-ence in heart disease diagnosis.This thesis provide a new way for establishing medical decision support system based on CBR, RBR method, and give a new direction for optimizing the efficiency of CBR case retrieval and improving the accuracy of RBR reasoning. We can make a conclusion that the medical decision support system for heart disease diagnosis based on the proposed model has some practical significance in improving the quality of medical care for heart disease.