Randomizing Outputs to Increase Prediction Accuracy
Report Number
518
Citation
Electronic Journal of Probability</em>, Vol. 5 (2000) Paper no. 2, pages 1-18
Abstract
Bagging and boosting reduce error by changing both the inputs and outputs to form perturbed training sets, grow predictors on these perturbed training sets and combine them. A question that has been frequently asked is whether it is possible to get comparable performance by perturbing the outputs alone. Two methods of randomizing outputs are experimented with. One is called output smearing and the other output flipping. Both are shown to consistently do better than bagging.
PDF File
Postscript File