Presentation 2015-03-06
Asymptotic Properties of Area Under the ROC Curve via Likelihood Ratio Based Ranking Function
Kentaro NAKANISHI, Toshiyuki TANAKA, Naonori UEDA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We study properties of the area under the ROC curve, which is one of measures to evaluate performance of a binary classifier constructed on the basis of a ranking function. We prove that the likelihood ratio is the optimal ranking function in the sense that it maximizes AUC when the probability distributions of the two classes are assumed to be known. We then study asymptotic properties of AUC via the likelihood ratio based ranking function by using maximum likelihood estimation of the model parameters, and numerical experiments are performed to confirm the asymptotic properties. We apply nonparametric estimation methods to the estimation of the likelihood ratio and perform numerical experiments to compare the asymptotic convergence rate with that via maximum likelihood estimation.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) AUC / Ranking Function / Asymptotic Properties / Density Ratio Estimation
Paper # IBISML2014-92
Date of Issue

Conference Information
Committee IBISML
Conference Date 2015/2/26(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-Based Induction Sciences and Machine Learning (IBISML)
Language ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Asymptotic Properties of Area Under the ROC Curve via Likelihood Ratio Based Ranking Function
Sub Title (in English)
Keyword(1) AUC
Keyword(2) Ranking Function
Keyword(3) Asymptotic Properties
Keyword(4) Density Ratio Estimation
1st Author's Name Kentaro NAKANISHI
1st Author's Affiliation Graduated School of Informatics, Kyoto University()
2nd Author's Name Toshiyuki TANAKA
2nd Author's Affiliation Graduated School of Informatics, Kyoto University
3rd Author's Name Naonori UEDA
3rd Author's Affiliation NTT CS-Labs.
Date 2015-03-06
Paper # IBISML2014-92
Volume (vol) vol.114
Number (no) 502
Page pp.pp.-
#Pages 8
Date of Issue