In order to take fully advantage of the diversity of classifier to improve the accuracy of classifiers ensemble, a classifier selection method is presented.The basic idea is to construct all base classifiers confusion matrix as cluster of data objects, and select a certain number of classification according to the distribution of the cluster sample as a representative, to be integrated to form a new collection of classifier.The method is applied to the training process Bagging algorithm, it is verified that the method can indeed improve the classifier ensemble performance through experiment.

multiple classifier systems; classifier selective ensemble; confusion matrix; cluster;