Classification of Thai Commercial Fish Sauces by Near-Infrared Spectroscopy with Chemometrics

Abstract

This study was aimed to find the suitable input wavelength variables and develop the models for classifying one hundred of Thai fish sauces by using near-infrared transflectance spectroscopy (NIR) with various chemometric methods. In the study, the wavelength interval selection methods named the Moving Window Partial Least Squares Regression (MWPLSR) and the Searching Combination Moving Window Partial Least Squares (SCMWPLS) were applied for searching suitable input wavelength variables. The methods were carried out by use of in-house-written program in MATLAB. Consequently, the Soft Independent Modeling of Class Analog (SIMCA) was used to develop the classification model. The results of suitable wavelength selecting were the absorbance regions at 1) spectra region at 1100-1900 and 2000-2440 nm²) informative region I at 1582-1762 nm selected by MWPLSR 3) informative region II at 2136-2428 nm selected by MWPLSR 4) direct combination of informative region I and II and 5) optimized combination of informative regions I and II at 2264-2428 nm selected by SCMWPLS. The developed classification models using SIMCA showed that five different input absorbance regions of selective wavelengths were able to classify fish sauces. All five models produced the corrective classification rate greater than 70%.