Supervised principal component analysis

Abstract

The invention provides a multivariate modeling method for quantitative analysis by supervised principal component analysis (SPCA). The method comprises: (a) designing a plurality of calibration samples wherein the desired variances are dominant or greatly enhanced; (b) producing a calibration data matrix using suitable mathematical pretreatment and truncation of the acquired NIR/Raman spectra of the calibration samples; (c) decomposing the matrix using PCA; (d) evaluating the score and loading matrices to ensure a genuine orthogonal relationship between scores of the desired latent variables in a two-dimensional principal component space 7; (e) generating a prediction matrix for quantitative prediction of unknown samples. This method does not require testing of calibration samples using a reference method. In addition, this method has high tolerance to variations in sample composition and manufacturing conditions.

Dependencies are inferred from the claim text.