Presentation 2011-10-20
Evaluating the Risk of Nonlinear Prediction with the Bagging Algorithm
Kazuya NAKATA, Tomoya SUZUKI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Some real phenomena are derived from unstationary systems, and therefore we have to select recent historical data which is not long enough as learning data to make a predictor. However, short learning data reduce the learning ability of predictors and often cause large prediction errors. Thus, in the previous study, we estimate the distribution of future states by the bagging predictors, and also estimate the risk of prediction errors according to the standard deviation of the estimated distribution. Moreover, we demonstrate that the estimated risk helps us to avoid dangerous predictions and to improve the prediction accuracy and reliability through computational simulations using some short data derived from chaotic models or real systems.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) short data / nonlinear prediction / ensemble learning
Paper # CAS2011-33,NLP2011-60
Date of Issue

Conference Information
Committee CAS
Conference Date 2011/10/13(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Circuits and Systems (CAS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Evaluating the Risk of Nonlinear Prediction with the Bagging Algorithm
Sub Title (in English)
Keyword(1) short data
Keyword(2) nonlinear prediction
Keyword(3) ensemble learning
1st Author's Name Kazuya NAKATA
1st Author's Affiliation Graduate School of Science and Engineering, Ibaraki University()
2nd Author's Name Tomoya SUZUKI
2nd Author's Affiliation Graduate School of Science and Engineering, Ibaraki University
Date 2011-10-20
Paper # CAS2011-33,NLP2011-60
Volume (vol) vol.111
Number (no) 242
Page pp.pp.-
#Pages 6
Date of Issue