Review of modern logistic regression methods with application to small and medium sample size problems

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

6 Citations (Scopus)

Abstract

Logistic regression is one of the most widely applied machine learning tools in binary classification problems. Traditionally, inference of logistic models has focused on stepwise regression procedures which determine the predictor variables to be included in the model. Techniques that modify the log-likelihood by adding a continuous penalty function of the parameters have recently been used when inferring logistic models with a large number of predictor variables. This paper compares and contrasts three popular penalized logistic regression methods: ridge regression, the Least Absolute Shrinkage and Selection Operator (LASSO) and the elastic net. The methods are compared in terms of prediction accuracy using simulated data as well as real data sets.

Original languageEnglish
Title of host publicationAI 2010
Subtitle of host publicationAdvances in Artificial Intelligence - 23rd Australasian Joint Conference, Proceedings
PublisherSpringer
Pages213-222
Number of pages10
ISBN (Print)3642174310, 9783642174315
DOIs
Publication statusPublished - 2010
Externally publishedYes
EventAustralasian Joint Conference on Artificial Intelligence 2010 - Adelaide, Australia
Duration: 7 Dec 201010 Dec 2010
Conference number: 23rd
https://link.springer.com/book/10.1007/978-3-642-17432-2 (Proceedings)

Publication series

NameLecture Notes in Computer Science
PublisherSpringer
Volume6464
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceAustralasian Joint Conference on Artificial Intelligence 2010
Abbreviated titleAI 2010
CountryAustralia
CityAdelaide
Period7/12/1010/12/10
Internet address

Keywords

  • Elastic Net
  • LASSO
  • Logistic regression
  • Ridge regression
  • Variable Selection

Cite this