A proposal of a new variable selection method utilizing gene's distribution information and solutions search progress rate of Real-coded genetic algorithms

Abstract

Recently variable selection and parameter optimization are getting more and more important. Regarding parameter optimization, much attention has been paid to Real-coded Genetic Algorithms (RCGA) because of their good searching ability and high flexibility. As for variable selection, traditionally Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC) are used quite often as selection criteria. These criteria estimate the relative quality of analysis models for a given set of data, but do not evaluate the importance of the variables themselves. This paper proposes a new variable selection method applying RCGA. This new variable selection method consists of 2 main components. The one is a new variable selection criterion utilizing the variances of genes in RCGA and the other is an estimation method of how far is in progress of RCGA optimization. The effectiveness of this new variable selection method is confirmed through application to a multiple linear regression model.