Abstract: SVM(Support Vector Machine) is an important tool of solving the nonlinear problem. This paper introduces the methods of constructing Support Vector Stepwise Regression speedily ----starting from the Multiple Linear Regression model of the Sample Subset to search the Support Vectors with the aim of decreasing the size of the Kernel Matrix and reducing the computing complexity of Support Vector Stepwise Regression, analyzes the complexity of the algorithm and illustrates an application example.