Application Of Support Vector Machine In Medical Data Analysis

Support Vector Machine (SVM) is an implementation of minimization principle of structure risk in statistics theory. Because of its excellent learning capability, this technology is becoming hot subject of research in machine learning and has been widely used in many fields, such as voice recognition, face detections, etc.A new algorithm has been proposed in this paper by combining modified FCM and SVM. By preprocessing training samples, extracting all possible support-vectors and removing outliers, the new algorithm improves the accuracy of classification and training speed.In this paper, the modified algorithm is used in medical data analysis. The main contents of this paper are outlined as follows:(1) Classify abnormal gait of children. By comparing three Classifiers using three different kernel functions, the conclusion is that the classifier with radial basis function as kernel is the best, with the accuracy of 97.51%. Moreover, by comparing the algorithm with others, the accuracy is better. So the result shows this method can effectively identify gait changes. It is a better solution to the problem of poor generalization of small sample of data in the classification of gait.(2) Classify three kinds of ECG. Firstly, the ECG is preprocessed. The mathematical morphology is used to remove baseline drift. Secondly, wavelet modulus maximum method is applied to detect R peak point. Thirdly, wavelet packet is used to extract energy feature. Finally, the classification accuracy can be achieved at 97.03% by taking the algorithm.(3) Diagnose the heart disease. The data from UCI machine learning library are classified using the conventional algorithm and the modified algorithm respectively. The result shows that modified algorithm is superior to conventional algorithm. In addition, comparing with other algorithms, the algorithm has better learning capability and generalization capability.As a whole, the modified algorithm has advantage in classification accuracy, feasibility and effectiveness. The excellent properties of this algorithm shows the great potential of application in medical diagnosis.