基于一个球的模式分类方法的改进

针对一个球的模式分类(Single Sphere Pattern Classification(SSPC))方法中选取参数C比较困难的问题,提出一种改进的分类方法υ-SSPC.这种方法通过引入一个具有明确物理意义的参数υ,即υ是间隔错误样本占所总样本点的分额的上界,是支持向量的个数所占总样本点数的分额的下界,使参数可以灵活地根据实际问题的精度要求来选取.从而可以快速选取最有效的参数,提高分类预测的精度.

A New Pattern Classification Method Based on Single Sphere

If we use Single Sphere Pattern Classification(SSPC),the difficulty in selecting the most effective error penalty has not been resolved.In this paper,we propose a υ-SSPC classification method by improving upon SSPC.This method provides the facility to counter these effects by introducing a parameter υ which has specific meanings that represent the upper bound for the the fraction of error samples of whole data sets,and the lower bound for the fraction of support vectors of whole data sets.As such this bound enable the training of machines with specific error rate requirements and the predictive accuracy will be improved.