Principal weighted logistic regression for sufficient dimension reduction in binary classification

Boyoung Kim, Seung Jun Shin

Research output: Contribution to journalArticle

Abstract

Sufficient dimension reduction (SDR) is a popular supervised machine learning technique that reduces the predictor dimension and facilitates subsequent data analysis in practice. In this article, we propose principal weighted logistic regression (PWLR), an efficient SDR method in binary classification where inverse-regression-based SDR methods often suffer. We first develop linear PWLR for linear SDR and study its asymptotic properties. We then extend it to nonlinear SDR and propose the kernel PWLR. Evaluations with both simulated and real data show the promising performance of the PWLR for SDR in binary classification.

Original languageEnglish
Pages (from-to)194-206
Number of pages13
JournalJournal of the Korean Statistical Society
Volume48
Issue number2
DOIs
Publication statusPublished - 2019 Jun 1

Fingerprint

Sufficient Dimension Reduction
Binary Classification
Logistic Regression
Reduction Method
Inverse Regression
Supervised Learning
Asymptotic Properties
Predictors
Data analysis
Machine Learning
kernel
Evaluation

Keywords

  • Binary classification
  • Model-free feature extraction
  • Weighted logistic regression

ASJC Scopus subject areas

  • Statistics and Probability

Cite this

Principal weighted logistic regression for sufficient dimension reduction in binary classification. / Kim, Boyoung; Shin, Seung Jun.

In: Journal of the Korean Statistical Society, Vol. 48, No. 2, 01.06.2019, p. 194-206.

Research output: Contribution to journalArticle

@article{99b6d332461f43aaa7ebbdcee06f36fb,
title = "Principal weighted logistic regression for sufficient dimension reduction in binary classification",
abstract = "Sufficient dimension reduction (SDR) is a popular supervised machine learning technique that reduces the predictor dimension and facilitates subsequent data analysis in practice. In this article, we propose principal weighted logistic regression (PWLR), an efficient SDR method in binary classification where inverse-regression-based SDR methods often suffer. We first develop linear PWLR for linear SDR and study its asymptotic properties. We then extend it to nonlinear SDR and propose the kernel PWLR. Evaluations with both simulated and real data show the promising performance of the PWLR for SDR in binary classification.",
keywords = "Binary classification, Model-free feature extraction, Weighted logistic regression",
author = "Boyoung Kim and Shin, {Seung Jun}",
year = "2019",
month = "6",
day = "1",
doi = "10.1016/j.jkss.2018.11.001",
language = "English",
volume = "48",
pages = "194--206",
journal = "Journal of the Korean Statistical Society",
issn = "1226-3192",
publisher = "Korean Statistical Society",
number = "2",

}

TY - JOUR

T1 - Principal weighted logistic regression for sufficient dimension reduction in binary classification

AU - Kim, Boyoung

AU - Shin, Seung Jun

PY - 2019/6/1

Y1 - 2019/6/1

N2 - Sufficient dimension reduction (SDR) is a popular supervised machine learning technique that reduces the predictor dimension and facilitates subsequent data analysis in practice. In this article, we propose principal weighted logistic regression (PWLR), an efficient SDR method in binary classification where inverse-regression-based SDR methods often suffer. We first develop linear PWLR for linear SDR and study its asymptotic properties. We then extend it to nonlinear SDR and propose the kernel PWLR. Evaluations with both simulated and real data show the promising performance of the PWLR for SDR in binary classification.

AB - Sufficient dimension reduction (SDR) is a popular supervised machine learning technique that reduces the predictor dimension and facilitates subsequent data analysis in practice. In this article, we propose principal weighted logistic regression (PWLR), an efficient SDR method in binary classification where inverse-regression-based SDR methods often suffer. We first develop linear PWLR for linear SDR and study its asymptotic properties. We then extend it to nonlinear SDR and propose the kernel PWLR. Evaluations with both simulated and real data show the promising performance of the PWLR for SDR in binary classification.

KW - Binary classification

KW - Model-free feature extraction

KW - Weighted logistic regression

UR - http://www.scopus.com/inward/record.url?scp=85058181620&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85058181620&partnerID=8YFLogxK

U2 - 10.1016/j.jkss.2018.11.001

DO - 10.1016/j.jkss.2018.11.001

M3 - Article

VL - 48

SP - 194

EP - 206

JO - Journal of the Korean Statistical Society

JF - Journal of the Korean Statistical Society

SN - 1226-3192

IS - 2

ER -