Improved predictive models by indirect classification and bagging for classification, regression and survival problems as well as resampling based estimators of prediction error.