基于MTS特征选择的神经网络集成方法

Feature Selection based on Mahalanobis-Taguchi System for Neural Network Ensemble Method

在Bagging算法基础上,运用马田系统进行特征选择,形成双重扰动改善神经网络集成的分类性能.实验表明,双重扰动增加了集成网络个体精度和差异度,基于MTS-Bagging算法的分类性能相比于Bagging有明显提高.

To improve the classification ability of ensemble neural networks,MahalanobisTaguchi System is used for features selection based on Bagging algorithm.Experiment results show that the double disturbance increase classification ability and difference of the individual network.The classification ability of proposed method outperforms Bagging algorithm.