摘要： 为提高集成分类器系统的分类性能，提出一种分类器选择集成算法MCC-SCEN。该算法选取基分类器集中具有最大互信息差异性的子集和最大个体分类能力的子集，以确定待扩展分类器集，选择具有较大混合分类能力的基分类器加入到待扩展集中，构成集成系统，进行加权投票并产生结果。实验结果表明，该方法优于经典的AdaBoost和Bagging方法，具有较高的分类准确率。

关键词: 多分类器系统, 选择集成, 差异性, 分类能力, 加权投票

Abstract: In order to improve the performance of Multiple Classifier System(MCS), a new classifier selection ensemble algorithm called MCC-SCEN is proposed. This algorithm first finds classifier subsets which have the biggest single accuracy and diversity to generate the unexpand classifier set, then selects classifiers that have largest MCS to construct the ensemble system and uses weighted vote as ensemble method. Experimental results show this method is better than AdaBoost and Bagging method.

Key words: Multiple Classifier System(MCS), selection ensemble, diversity, classification capability, weighted vote

中图分类号: 

• TP312