Presentation 2014-07-18
A Prediction Method based on Weighted Ensemble of Decision Tree on Alternating Decision Forests
Shotaro MISAWA, Naohiro FUJIWARA, Kenta MIKAWA, Masayuki GOTO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this study, we focus on the Alternating Decision Forests (ADF). The ADF introduces the weights which represent the degree of prediction accuracy for the training data. These weights are only used when growing decision trees in the learning phase to improve the predictability for all training data. By using the weights in the prediction phase, it is possible to construct a prediction considering prediction confidence of each decision tree's leaf node. Moreover, by pruning the decision trees, the generalization ability can also be improved. Therefore, we propose the method introducing AIC criterion to prune brunches of the trees and using the weights when predicting the category label of a new input data. Our proposal can be interpreted as the predictive method to weaken the influence of the low confidence outputs and strengthen that of the high reliable outputs.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Data mining / Alternating Decision Forests / Decision Tree / ensemble
Paper # IT2014-29
Date of Issue

Conference Information
Committee IT
Conference Date 2014/7/10(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 Information Theory (IT)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A Prediction Method based on Weighted Ensemble of Decision Tree on Alternating Decision Forests
Sub Title (in English)
Keyword(1) Data mining
Keyword(2) Alternating Decision Forests
Keyword(3) Decision Tree
Keyword(4) ensemble
1st Author's Name Shotaro MISAWA
1st Author's Affiliation Graduate School of Creative Science and Engineering, Waseda University()
2nd Author's Name Naohiro FUJIWARA
2nd Author's Affiliation Graduate School of Creative Science and Engineering, Waseda University
3rd Author's Name Kenta MIKAWA
3rd Author's Affiliation School of Creative Science and Engineering, Waseda University
4th Author's Name Masayuki GOTO
4th Author's Affiliation School of Creative Science and Engineering, Waseda University
Date 2014-07-18
Paper # IT2014-29
Volume (vol) vol.114
Number (no) 138
Page pp.pp.-
#Pages 6
Date of Issue