COMPARISON OF MULTIVARIATE METHODS BASED ON LATENT VECTORS AND METHODS BASED ON WAVELENGTH SELECTION FOR THE ANALYSIS OF NEAR-INFRARED SPECTROSCOPIC DATA

Comparison of several calibration methods (principal component regress ion (PCR), partial least-squares, multiple linear regression), with an d without feature selection, applied on near-infrared spectroscopic da ta is presented for a pharmaceutical application. It is shown that PCR with selection of principal components instead of the usual top-down approach yields simpler and better models. As feature selection method s, selection of wavelengths correlated with concentration, with large covariance with concentration, with high loadings on the important pri ncipal components, and according to a method proposed by Brown, are co nsidered. The presented results suggests that feature selection can im prove multivariate calibration.