Because of the complexity of near infrared spectral data, effective strategies are necessary proposed for accurate quantitative analysis purpose. This work explores a new self-construction strategy for the arrangement of conventional near infrared two-dimensional spectra into new self-constructed three-dimensional spectra, and investigate the feasibility of N-way partial least squares combined with the new self-constructed three-dimensional near infrared spectra for obtaining accurate quantitative determination results. A proof-of-concept model system, the quantitative analysis of four components (moisture, oil, protein, and starch) in corn samples, was applied to evaluate the performance of the proposed strategy. The ability of the newly proposed approach to predict the target compounds was checked with test samples. The established models have good predictive power for the target compounds with acceptable values of Rp (range from 0.82 to 0.997) and RMSEP (range from 0.03 to 0.47). Compared with partial least squares method on pretreated near infrared spectra and N-way partial least squares method on the basis of near infrared self-constructed three-dimensional spectra, the proposed method is competitive.
A Modified Moving-Window Partial Least-Squares Method by Coupling with Sampling Error Profile Analysis for Variable Selection in Near-Infrared Spectral Analysis
