NIR Quantitative Model Optimization of Fatty Acid in Edible Oil Based on CARS-PLS

Abstract:

Competitive adaptive reweighted sampling (CARS) method was employed to improve the prediction accuracy of the NIR quantitative model of four kinds of fatty acid (palmitic acid, stearic acid, oleic acid and linoleic acid) in edible oil．Predict concentration residual method was employed to detect the outlier before preprocessing the spectroscopy by normalization．The key variables were selected by CARS method．The partial least squares (PLS) calibration models of four kinds of fatty acid were established respectively in the optimal conditions，and compared with the results using OPUS software. Determination coefficient (R2 )，root mean square error of cross validation(RMSECV)and root mean square error of prediction(RMSEP)were used to evaluate the quality of the modes．The results showed that better prediction was obtained by CARS. The result showed that using CARS could effectively simplify the model and the less number of wavelength variables selected could be reference for developing filter spectrometer of edible oil. 