A Diversity Measure for Tree-Based Classifier Ensembles

A Diversity Measure for Tree-Based Classifier Ensembles
复制标题

DOI:
10.1007/3-540-28397-8_4
复制
发表时间:
2005
期刊:
--
影响因子:
--
通讯作者:
Eugeniusz Gatnar
Eugeniusz Gatnar
中科院分区:
其他
文献类型:
--
作者:
Eugeniusz Gatnar

文献摘要

被引文献

相似文献

在过去的十年中,将多个分类器组合成一个集合被证明是非常成功的。这一成功的关键在于成分分类器的多样性,因为许多实验表明,不相关的成员组成了一个高精度的集合。本文提出了一种新的基于Hamann相似系数的分类器组合多样性的两两度量方法。
Combining multiple classifiers into an ensemble has proved to be very successful in the past decade. The key of this success is the diversity of the component classifiers, because many experiments showed that unrelated members form an ensemble of high accuracy.In this paper we propose a new pairwise measure of diversity for classifier ensembles based on Hamann’s similarity coefficient.