基于模糊系数规划的模糊支持向量分类机

Fuzzy support vector classifier based on fuzzy coefficient programming

本模糊支持向量分类机的构建特点是,训练点输出的类型和最终的模糊分类函数的函数值均为反映其模糊类别的实数。以模糊系数规划为基础,将模糊分类问题转化为求解模糊系数规划问题,求出模糊系数规划的γ-最优规划,据此给出模糊支持向量分类机(算法);用2个例子说明该算法的合理性;最后给出模糊支持向量分类机中最佳阈值的确定方法。

This paper is concerned with a fuzzy support vector classifier in which the type of the figures for both the output of the training point and the value of the final fuzzy classification function is real number. The characteristic of real number is propitious to indicate the class of the figures as either positive or negative one. First, a fuzzy classification problem was formulated as a fuzzy coefficient programming problem. Then this programming was transformed into its optimal programming. As a result, we proposed a fuzzy support vector classification algorithm. In order to show its rationality, two examples were presented. In addition, we also proposed a strategy to decide the optimal threshold value in our algorithm.