This article is cited in 4 papers

Regression models in the classification problem

V. M. Nedel'ko

Sobolev Institute of Mathematics, 4 Koptyug av., 630090 Novosibirsk

Abstract: We carry out a comparative analysis of the efficiency of several methods of classification (pattern recognition) based on regression models, in particular, logistic regression and its modifications. A new method for constructing a decision function is proposed. The method is based on the maximization of the area under the error curve over the linear functions in the space of the decision functions obtained by a special transformation. The performance of the method is illustrated by solving an applied problem.

Keywords: regression analysis, pattern recognition, machine learning, decision function, misclassification probability, logistic regression, Fisher linear discriminant.

UDC: 519.246

Received: 17.10.2013

Bibliographic databases:

•