2007 Volume 8 Pages 41-49
In this paper, independent component analysis (ICA) and regression analysis are combined to extract significant components. ICA is a method that extracts mutually independent components from explanatory variables. We propose a new method that selects combination of independent components by using genetic algorithm (GA). It can construct a PLS model that has high predictive accuracy. This method is named ICA-GAPLS. In order to verify the superiority of ICA-GAPLS, this method was applied to QSPR analysis of aqueous solubility. The result of comparison with PLS and other regression methods is shown. R2, Q2 and Rpred2 values of the PLS model are 0.826, 0.821 and 0.790, respectively. These values of the ICA-GAPLS model are 0.945, 0.882 and 0.889, respectively. ICA-GAPLS achieved higher predictive accuracy than PLS. ICA-GAPLS showed better result regarding Q2 and Rpred2 value than other methods. ICA-GAPLS could extract effective components from explanatory variables and construct the regression model having high predictive accuracy.