Bagging Predictors-英文文献.pdf

Bagging Predictors By Leo Breiman* Technical Report No. 421 September 1994 *Partially supported by NSF grant DMS-9212419 Department of Statistics University of California Berkeley, California 94720 Bagging Predictors 1 L eo Br eiman Department of Statistics University of California at Berkeley Abstract Bagging predictors is a metho d for generating multiple versions of a pre dictor and using these to get an aggregated predictor The aggregation av erages over the versions when predicting a numerical outcome and do es a plurality vote when predicting a class The multiple versions are formed by making b o otstrap replicates of the learning set and using these as new learning sets Tests on real and simulated data sets using classication and regression trees and subset selection in linear regression show that bagging can give substantial gains in accuracy The vital element is the instability of the prediction metho d If p erturbing the learning set can cause signicant changes in the predictor constructed then bagging can improve accuracy Intro duction A learning set of L consists of data fy x n N g where the y s n n are either class lab els or a numerical resp onse We have a pro cedure for using this learning set to form a predictor x L if the input is x we predict y by x L Now supp ose we are given a sequence of learnings sets fL g