An Introductory Study on

Abstract

In this study, we mainly dealt with the introduction and comparison of two optimization techniques over the estimated parameters and model results of the Binary Logistic Regression (LR) model. These are the traditional Newton-Raphson (NR) algorithm which requires the differentiable objective function and appropriate starting values related to the parameters and Genetic Algorithm (GA) approach which does not need any strict assumptions as the algorithm NR. The results suggest that NR algorithm and GA give very similar results when the assumptions of NR are satisfied. This indicates that GA could successfully be used instead of the NR method considering its more flexible assumptions. Moreover, if the objective function does not satisfy the differentiability condition, NR could not be used in the optimization process and fails to find the optimum values. Therefore, this study actually reveals the success of GA in the parameter estimation of LR model. All the model outputs are compared in terms of the estimated parameter values, ease of application and convergence rates. Matlab commands are also given with their explanations for GA for researchers studying in this area.