PLS and MLR methods in wavelength selection for multicomponent spectrophotometric data: a comparative study

Comparison of two calibration methods, multiple linear regression (MLR) and partial least squares), (PLS), with feature selection applied on UV-vis sp ectrophotometric data is presented. As feature selection methods, selection of wavelengths correlated with concentration (for MLR) and with high loadi ngs on the first factor, so as with high regression coefficients (bw) to th e model with the optimum number of factors (for PLS) are considered. For ea ch selected model, cross-validation is performed and the root mean squared prediction error (RMSECV) is calculated. It is shown that the PLS method yi elds better predictive ability (RMSECV) than the MLR method.