2011 Volume 10 Issue 4 Pages 122-130
Regions of explanatory variables X need to be selected in many fields such as spectral analysis and process control. Genetic algorithm-based wavelength selection (GAWLS) method is one of the methods that is used to select combinations of important variables from X-variables using regions as a unit of measurement. However, a partial least-squares method is used as a regression method; hence, a GAWLS method cannot handle a nonlinear relationship between X and an objective variable y. We therefore proposed a region selection method based on GAWLS and support vector regression (SVR), one of the nonlinear regression methods, for achieving both appropriate selection of variable regions and construction of a high predictive model when there is a nonlinear relationship between X and y(Figure 1). The proposed method is named GAWLS-SVR. The q2 value of a SVR model, which is calculated using a cross-validation method, is used as a fitness value of the chromosome. In order to verify the effectiveness of the GAWLS-SVR method, we applied it to simulation data in which correlation between close pairs of X-variables was high and the relationship between X and y was nonlinear. The GAWLS-SVR method could select regions of variables appropriately, while considering the nonlinearity and could construct a predictive model with high accuracy (Table 2, Figure 6).
