Study Of Support Vector Machine And Its Application In Cancer Diagnoses

Support Vector Machine is a new pattern recognition technology which is based on VCdimension theory of Statistical Learning Theory and the principle of Structural RiskMinimization. SVM can obtain the optimum result from the finite samples which is not onlythe optimum result when the samples are infinite. SVM has better generalization performanceand higher prediction accuracy to test example. So SVM has had a lot of application.Currently, SVM is becoming a hot area in the field of machine learning in the world.In view of its better learning capability and perspective in application, this paper tried toapply SVM to the cancer disease diagnoses. The main works are as follows:Firstly, Statistic Learning Theory and the theoretic base of SVM are introduced briefly.Secondly, an overview on a variety of classification algorithms for support vectormachine is given. Some algorithms such as C-SVM,ν -SVM, BSVM and LS-SVM are aimedto be discussed.Thirdly, SVM nonlinear classifier is employed to breast cancer disease diagnoses byoptimizing parameters of training models. High recognition rate is obtained in the prediction.Finally, on the base of the traditional SVM algorithm, the SVM is initially trained fromthe training samples, thereby producing a number of main support vectors. Subsequently, theSVM are re-trained only from the main support vectors, thereby producing a number ofsub-support vectors. In the area of cancer disease diagnoses, the SVM nonlinear classifierwhich is constructed by sub-support vectors, can achieve more higher recognition rate thanusing the traditional SVM classifiers in the prediction.