PCA Based Bagging Ensemble Methods

Redundant features in data sets hurt the generalization performance of learning machines.Even the state-of-arts algorithms such as support vector machine(SVM) and ensemble learning are not immune to it.This paper studies the effect of feature transformation with principal component analysis(PCA) on bagging method of ensemble learning.This paper proposes a new method termed PCA-bagging,which is compared with other methods such as single SVM,bagging of SVM and SVM with PCA.Experiments on UCI machine learning benchmark data sets show that PCA-bagging has better generalization performance,indicating that even ensemble learning methods with excellent generalization ability also need proper feature preprocessing.