一种基于证据距离的多分类器差异性度量

 多分类器系统因其能够显著提升分类精度而引发了广泛关注。多分类器系统中各子分类器间的差异性是提升融合分类精度的先决条件。提出了一种基于证据距离的分类器系统差异性度量,同时基于该度量提出一种多分类器系统构造方法。综合了既有差异性度量、所提新差异性度量以及在训练样本集上的分类性能等多个指标,实现了多分类器系统的有效构造。实验结果表明,所提差异性度量及多分类器系统构造方法是合理的,能有效提升融合分类精度。

A Novel Diversity Measure of Multiple Classifier Systems Based on Distance of Evidence

Multiple classifier systems can effectively improve the classification performance in many applications,which is why they have attracted a great deal of interest.Diversity among member classifiers is a necessary condition for improvement in classifier ensemble performance.In this paper,a novel diversity measure of multiple classifier systems is proposed based on the distance of evidence and a new approach to multiple classifier system design is presented.By using jointly the proposed diversity measure,the traditional diversity measure and the classification performance on training samples,an effective multiple classifier system can be implemented.It is experimentally shown that the proposed diversity measure and the proposed approach to multiple classifier system design are rational and effective.