A Self-Organizing Model for Logic Regression

Stanley Jerry Farlow

Abstract

Logic regression, as developed by Ruczinski, Kooperberg, and LeBlanc (Ruczinski, Kooperberg, and LeBlanc 2003) is a multivariable regression methodology that constructs logical relationships among Boolean predictor variables that best predicts a Boolean dependent variable.  More specifically, they find a regression model of the form g(E/Y)= b0+b1L1+...+bmLm  where both the coefficients b0,b1,...,bm and the logical expressions Lj, j=1,...,m  are determined.  The logical expressions  are logical relationships among the predictor variables, such as "X1,X2 are true but not X5"  , or "X3,X5,X7 are true but not X1 or X2".  In their paper, the authors investigate the use a simulated annealing algorithm.  In this paper, we use the Group Method of Data Handling (GMDH) to approach the problem.

 

 

Keywords

Logic Regression, GMDH algorithm, Self-organizing methods

Full Text:

PDF