The Research On SVM Multi-classification Forecast

Support Vector Machine(SVM) is a novel powerful machine learning method roted in the framework of Statistical Learning Theory(SLT) .SVM has become a new active area in the field of forecast becase of its excellent learning performance .The theory and applications of two-class classification problem have gradually come to maturity,while the research of multi-class classification problem is weak either in scope or in depth.At present,many serious accidents about coal and gas outburst frequently happened,which jeopardized to peoples lives and wealth .The process of outburst is highly complicated,a forecast of coal and gas outburst has many characteristics ,such as various influent-factors ,higher-dimension nonlinear and so on .This paper applies SVM methord to forecast of coal and gas outburst .It has actual significance .The contributions have rade lets of literatures (both English and Chinese),the knowledge of coal and gas outburst is established .The correlation between outburst and other variables is analyzed .Besides,a review and detail descriptions of SVM are also shown in this paper .After it analyzes and compares several techniques of SVM,it ascertains the suitable techniques required in the forcast of coal and gas outburst,and explain the necessary theory to them .This paper studies and compares the performance of several multi-class classification methods based on SVM theory .Next it introduces Gauss kernel fuction and discusses its separability and the property of localization .It also considers feature selection method based on the sequential minimization technique .As a resalt it can improve SVM application ability in classification problem .Established a model .According to the circumtemces of physical geography in the selected area of coal mine ,it ascertains input factors,and selects suitable kernel fuction and parameters in the forecast .This paper establishes a model during the course of the forecast by using SVM .This experiment obtains better effect in accordance with procedure in this forecast .