Research And Application Of Some Support Vector Machine Algorithm

Support vector machine (SVM) is a kind of new method, new technology in thefield of data mining. It is suitable for pattern recognition with small sample,nonlinear and high dimensional;It is mainly to solve the classification and regressionproblems. Because of its good generalization ability, SVM achieved good results inthese aspects of support vector machine. It is based on the largest interval andestablish the optimal hyperplane in the original space for data classification orregression. Sometimes, it is need to introduce the kernel function,and mapping theoriginal data to a high-dimensional feature space; then establish the optimal hyperplane in the feature space to solve the classification and regression problems.The main work of this paper are as follows:The first chapter introduced the background and significance of support vectormachine research, support vector machine status, this research topic origin and themain research content;The second chapter introduces the theoretical basis of support vector machine,including the statistical learning theory, the core idea of support vector machine,typical classification support vector machine and its variant;The third chapter using Mahalanobis distance instead of Euclidean distance,launched four kinds of kernel function based on the Mahalanobis distance, andreformed six kinds of classic support vector classification;The fourth chapter by Fisher discriminant theory, introduced the four class "F"kernel function;Then structured all kinds of support vector machine based on theone kind of Fisher discriminant;The fifth chapter combined the fuzzy maximum membership degree and attributemeasure criterion are combined, and improved the attribute measure criterionmodel; then use the model to construct six kinds of support vector classificationbased on a class of attribute measurement.The sixth chapter data collected section of river water quality in China, and usethe the above24kinds of support vector to analysis; then compare all kinds ofimproved support vector machine with the classical support vector machine is goodor bad. The results found,support vector machine based on Mahalanobis distance,support vector machine based on a class of Fisher discriminant and support vectormachines based on one class attribute measure achieved good classification results.Some classification support vector machine is better than the classical effect.