Prediction of Edible-Oil Acid Values and Identification of Oil Species Based on Near Infrared Spectroscopy

Quantitative prediction of acid value and qualitative identification of edible oils were studied on the basis of near infrared spectroscopy. Four preprocessing methods including multivariate scattering correction (MSC), combination of standard normal variate and de-trend (SNV-DT), moving average smoothing (MAS), and Savitzky-Golay (SG) were used. Successive projection algorithm (SPA), interval partial least squares (iPLS), combination of competitive adaptive reweighted sampling algorithm and partial least squares method (CARS-PLS) were applied in the extraction of characteristic wavelengths. Particle swarm optimization (PSO) and genetic algorithm (GA) were used to establish a variety of support vector machine (SVR) models for the quantitative prediction of acid values. According to the prediction results of these models, the optimal technique was selected.