基于十折法的最小二乘支持向量机参数选取方法

Parameters Selection Method Based on Ten-fold Least Squares Support Vector Machines

针对最小二乘支持向量机参数选取的问题,提出一种基于十折法的参数选取方法。将数据平均分成十组,采用一组作为训练特征,其它作为测试组,规定参数的可能范围,然后选取平方差最小的作为参数的取值。经实验证明结果令人满意,提出的方法是可行的。

Parameters selection method based on ten-fold least squares support vector machines is presented. The data are divided into ten groups, using a set of characteristics as the training, the other as the test group, the provisions of the parameters possible, and then the smallest square difference value is selected as a parameter. The experimental results is satisfactory. The proposed method is feasible.