关键词: 最小二乘支持向量机, 类内散度, 更好精度, 线性系统

Abstract: Support Vector Machine（SVM） is one of focuses of research and application in classification.A new least-squares-based algorithm that introduces a within-class scatter with guaranteed classification performance（VSLSVM） in the design of least squares support vector machines（LS-SVM） is presented.This algorithm can obtain better correctness that reformulates primal LS-SVM problems with optimality criterion Min w′Mw where w is the weight vector corresponding the primal LS-SVM problems，M is the within-class scatter matrix.This method only requires to solve a linear system instead of a quadratic programming problem.Experiments are included to compare SVM and Suykens’ approach.

Key words: east squares support vector machines（LS-SVM）, with-class scatter, better correctness, linear system