Genetic algorithm based neural classifiers for factor subset extraction

Abstract

A clear understanding of risk factors is important to develop appropriate prevention and control strategies for infection caused by such pathogens as Salmonella Typhimurium. In this study, there are 91 risk factors that nonlinearly contribute to the Salmonella Typhimurium infection and many of them are not of significance. It is very important to automatically extract a factor subset with those important risk factors. This paper proposes a genetic algorithm for factor subset extraction in conjunction with neural and statistical classifiers to classify case and control status in Salmonella Typhimurium infection. The results show that the proposed approach is able to find an appropriate factor subset and the proposed neural classifiers outperform the traditional statistical classifiers. A statistical analysis is conducted by varying the parameters in the genetic algorithm based neural classifier to minimise the prediction error and determine the optimal system configuration.