Artificial intelligence

Scientific journal

ISSN 2710-1673

ONLINE: ISSN 2710-1681

Select your language


Відбір змінних в логістичну регресію генетичним алгоритмом

Paklin N.1
1 GOU VPO "Ryazan Branch of the Moscow State University of Economics, Statistics and Informatics"

Full text (PDF)

UDC: 62-50:15
Publication Language: Russian
Stuc. intelekt. 2008; 13(3): 714-719

Abstract: In the paper we discuss effective procedures for а feature selection problem in a binary logistic regression model. A genetic algorithm was used to find best feature combinations, with the special fitness function based on a penalty parameter for including new variables. This parameter depends on ROC-curve index on current epoch. Experiments on Madelon data set and credit scoring classification problem were made.

Keywords:

References:

View full text (PDF)