基于ECOC的多类代价敏感分类方法

Multiclass Cost-sensitive Classification Based on Error Correcting Output Codes

研究了基于纠错输出编码实现多类代价敏感分类的方法,提出了一种新的将多类代价敏感分类问题分解为多个二类代价敏感分类问题的框架。为获得其中每个二类代价敏感基分类器的二类代价矩阵,提出了利用已知多类代价矩阵计算误分类代价的期望值的方法,给出了计算二类代价矩阵的通用计算公式。为验证所提方法的有效性,在人工和UCI数据集上将其与现有方法进行了比较,实验结果表明所提方法具有相似甚至更好的性能。

Approach of multiclass cost-sensitive classification based on error correcting output codes is studied in this paper,and a new framework to decompose the complex multiclass cost-sensitive classification problem into a series of binary cost-sensitive classification problems is proposed.In order to obtain the binary cost matrix of each binary cost-sensitive base classifier,a method of computing the expected misclassification costs from the given multiclass cost matrix is proposed,and the general formula for computing the binary costs are given.Experimental results on artificial datasets and UCI datasets show that the proposed method has similar or even better performance in comparison with the existing methods.