Abstract: A near infrared (NIR) spectroscopic model for the quantitation of flour moisture was developed using partial least squares regression (PLSR). Infrared spectra of 130 randomly selected samples were used to establish PLSR models employing different spectral pretreatments combined with wavelength selection using genetic algorithm (GA). The correlation coefficient (R2), root mean square error of calibration (RMSEC) and root mean square error of prediction (RMSEP) of the optimized model were 0.977 7, 0.245 3 and 0.264, which were respectively higher, lower and lower than those of the unoptimized one. Thus tt is feasible to establish a quantitative model for estimating flour moisture by spectral data pretreatment and GA-based wavelength selection which had better accuracy and lower errors than the unoptimized one.

Key words: spectral data pretreatment, genetic algorithm (GA), near infrared spectroscopy, partial least squares (PLS), flour, moisture