Seguimiento en modelos de regresión logística

Bravo, Cristian and Maldonado, Sebastian (2009) Seguimiento en modelos de regresión logística. Revista Ingeniería Industrial, 8 (2), 31-44.

Record type: Article

Abstract

Most data mining projects in real life applications give as a result only static solutions which, in time, lose their inherent capacity to explain the phenomena they were originally built for. We introduce an theoretical-practical framework that allows to closely follow up logistic regression models to determine the moment when they must be updated, maintaining an strict control over their evolution, the variables in them and relevant changes that can occur in the population since they were originally designed. The statistical test presented include classical tests such as Kolmogorov-Smirnov and Chi-Squared statistic to measure changes in means of the variables present in the models, plus a novel test designed from the distribution of the models coefficients that allows to measure the moment when a population has changed more than the confidence intervals defined from the original parameters. The methodology was tested using the databases from two real world micro-entrepreneurs credit scoring projects developed between the years 2007 and 2008, with very good results

This record has no associated files available for download.