Selective SVM Ensemble Based On Fuzzy Clustering

Support Vector Machine (SVM) is a kind of new machine learning algorithm, its theories foundation is the covariance study theories that the Vapnik establish. Support Vector Machine (SVM), which is based on Structural Risk Minimization (SRM) and the Functions of kernel, has shown much better performance than most other existing machine learning methods. In recent years, with the emergence of the training algorithm quickly, its application research such as face detection, verification and identify handwriting, speaker/speech identification, character/script identification, image processing, and other applications.About the application of support vector machine in remote sensing image, the work and contribution of this article is followed:(1) have taken part in the program research which is called technique research of pattern recognition of forest resources which is based on remote sense of high spectrum;(2) introduce the theory that is selective support vector machine ensemble based on fuzzy clustering, which is raised after several times discussion. First we train several support vector machines use Bootstrap method, then assemble them to select the one closest to the cluster center, at last selected ones are assembled together and classify by ballot;(3) the proposed algorithm has realized in MATLAB. The experiment result prove the feasibility of this algorithm, which is the classification of ensemble SVM is better than individual one.(4) according to the principle of clustering and support vector machine, the article also proposed that first clustering original data, then select centers of all clusters to train the support vector machine. The results indicate that this method can gain the better classification accuracy.