Optimal Determination of K Constant of Ridge Regression Using a Simple Genetic Algorithm

In the present work, the optimal determination of the constant k of ridge regression (RR) model was developed, since this method of regression permit to reduce the multicollinearity problem and it is a one advantage that takes the ridge regression model over ordinary least square (OLS). The optimal constant k was developed by some technique derived from intelligent systems (genetic algorithm) and some statistics techniques. In this paper we present a comparison between these two methodologies for finding the optimal constant k since this constant givesless variance, giving stability to the estimate coefficients and reduce the multicollinearity problem. The results analyzed by the statistical methods, showed that MSR and R² have very good performance but theVIF´s are greater than genetic algorithm (GA) and theGA reduces the VIF´s but reduces R² and increment the MSR.