Research On Application Of Rough Set Attribute Reduction Methods In Medical Diagnosis

Now the hospital has accumulated vast amounts of medical diagnostic data, utilization of a large amount of data using advanced information processing technology has become an important direction of the development of medical. At present the method of medical diagnosis is mainly based on the diagnosis to the patient’s symptoms. But with the increasing of the types of diseases, the interference of symptoms is greatly enhanced, which brings great burden to the doctor. In this paper, through the analyzing of the problem of medical diagnostic and the attribute reduction algorithm based on rough set, two kinds of attribute reduction algorithm is proposed for medical diagnosis. The main work is as follows:1) In view of the problem that the same weight of the attributes is often occurred in the results obtained by the HORAFA algorithm and other existing improved algorithm. According to the features of medical diagnostic data, a new heuristic reduction algorithm based on discernibility matrix is put forward by improving the heuristic rules and the attribute deletion operation of the algorithm. The experimental results show that the algorithm can improve the efficiency of reduction and obtain better reduction.2) A new kind of adaptive genetic reduction algorithm is put forward in order to solve the attribute reduction problem of large-scale medical diagnostic data. The improved attribute weights is used to construct individual fitness function. The new optimal selection strategy is used to enrich the types of community, to avoid local extremum. The concept of attribute similarity is used to reduce the redundant operation of crossover and the calculation number of fitness function. Improving the mutation operation is to avoid that attributes with the same weight exist in each individual. Experiments show that the algorithm is more suitable for the information system which has a large amount of data.3) Genetic reduction algorithm is applied to the reduction of the 2 diabetes mellitus data, to extract the key symptoms, to assist process of diagnose, and to reduce the rate of misdiagnosis and missed diagnosis...