| Summary: | 碩士 === 銘傳大學 === 電子工程學系碩士班 === 98 === This paper will propose the Analysis of heart disease database and the design of automatic diagnostic heart disease systems. In the beginning, we introduce the heart disease database and rank the attributes in this database according to their importance. These attributes contain two kinds of data, which are category attributes and numerical attributes. Numerical attributes have information that are measured by many medical equipments such as a typical electrocardiograph machine. Category attributes are defined by their attribute value：The ECG abnormalities can be defined as one and the normal ECG can be defined zero. Then we use the methods of grey statistics and some data-mining technology to order the attributes in sequence. To get higher classification ratio by excluding the lower important attributes. The goal of automatic diagnostic heart disease systems is to classify patients correctly. Our proposed methods are related to Grey Relation, Cerebellar Model Articulation
Controller, and Decision-Tree, respectively. In Grey Relation method, we find that the correct ratio can be attained 94.44%, when the Gauss function is embedded into the traditional grey relation function. In Cerebellar Model Articulation Controller class, we can get 90% correct ratio by using two attributes that have higher important level and the secondary choose by other attributes. In Decision-Tree method, we set the threshold value that suggested by doctors’ experience to separate data into binary. We find that only the correct ratio 83.33% can be derived using some integrated attributes. The reason may be inadequate in data number or the threshold values which are not
fine enough. This is the way to improve in future.
|
|---|
