A Robust Bagging Method Using Median as a Combination Rule

Bagging has been known to be successful in increasing the accuracy of prediction of the unstable classifiers. In bagging predictors are constructed using bootstrap samples from the training sets and then aggregated to form a bagged predictor. The robust bagging discard the bootstrapped classifiers generating extreme error rates, as estimated by the out-of-bag error rate and to combine over the remaining ones using the robust location estimator,´median´. In this paper we try to explore the advantages of robust bagging. We carried out experiments on several benchmark data sets and suggest from the results that robust bagging performs quite similar compare to the standard bagging when applied to unstable base classifiers such as decision trees, but performs better when applied to more stable base classifiers as Fisher linear discriminant analysis and nearest mean classifier.