Header

UZH-Logo

Maintenance Infos

On the choice and influence of the number of boosting steps


Seibold, Heidi; Bernau, Christoph; Boulesteix, Anne-Laure; De Bin, Riccardo (2016). On the choice and influence of the number of boosting steps. Department of Statistics: Technical Reports: LMU.

Abstract

In biomedical research, boosting-based regression approaches have gained much attention in the last decade. Their intrinsic variable selection procedure and their ability to shrink the estimates of the regression coefficients toward 0 make these techniques appropriate to fit prediction models in the case of high-dimensional data, e.g. gene expressions. Their prediction performance, however, highly depends on specific tuning parameters, in particular on the number of boosting iterations to perform. This crucial parameter is usually selected via cross-validation. The cross-validation procedure may highly depend on a completely random component, namely the considered fold partition. We empirically study how much this randomness affects the results of the boosting techniques, in terms of selected predictors and prediction ability of the related models. We use four publicly available data sets related to four different diseases. In these studies the goal is to predict survival end-points when a large number of continuous candidate predictors are available. We focus on two well known boosting approaches implemented in the R-packages Cox-Boost and mboost, assuming the validity of the proportional hazards assumption. Finally, we empirically show how the variability in selected predictors and prediction ability of the model is reduced by averaging over several repetitions of cross-validation in the selection of the tuning parameters.

Abstract

In biomedical research, boosting-based regression approaches have gained much attention in the last decade. Their intrinsic variable selection procedure and their ability to shrink the estimates of the regression coefficients toward 0 make these techniques appropriate to fit prediction models in the case of high-dimensional data, e.g. gene expressions. Their prediction performance, however, highly depends on specific tuning parameters, in particular on the number of boosting iterations to perform. This crucial parameter is usually selected via cross-validation. The cross-validation procedure may highly depend on a completely random component, namely the considered fold partition. We empirically study how much this randomness affects the results of the boosting techniques, in terms of selected predictors and prediction ability of the related models. We use four publicly available data sets related to four different diseases. In these studies the goal is to predict survival end-points when a large number of continuous candidate predictors are available. We focus on two well known boosting approaches implemented in the R-packages Cox-Boost and mboost, assuming the validity of the proportional hazards assumption. Finally, we empirically show how the variability in selected predictors and prediction ability of the model is reduced by averaging over several repetitions of cross-validation in the selection of the tuning parameters.

Statistics

Downloads

12 downloads since deposited on 12 Jan 2017
12 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Scientific Publication in Electronic Form
Communities & Collections:04 Faculty of Medicine > Epidemiology, Biostatistics and Prevention Institute (EBPI)
Dewey Decimal Classification:610 Medicine & health
Language:English
Date:2016
Deposited On:12 Jan 2017 09:35
Last Modified:14 Sep 2017 13:14
Publisher:LMU
Free access at:Official URL. An embargo period may apply.
Official URL:https://epub.ub.uni-muenchen.de/26724/

Download

Download PDF  'On the choice and influence of the number of boosting steps'.
Preview
Filetype: PDF
Size: 472kB