Fig. 1. Raw Fourier transform infrared (FTIR) spectra for pure sesame and cottonseed oil.
Fig. 2. Principal component analysis (PCA) score plot for sesame oil (a) and cottonseed oil (b) after applying multiplicative scatter correction (MSC) preprocessing.
Fig. 3. First three principal components loading plots from principal component analysis (PCA) for benzene-adulterated sesame oil (a), (b), (c) and the spectrum of pure benzene with variable importance in projection (VIP) and selectivity ratio (SR) for selected variables (d).
Fig. 4. Regression plot of the actual versus calculated percentages of benzene in the validation set of the whole spectral region (a), VIP (b) and SR (c) variable selection methods. VIP, variable importance in projection; SR, selectivity ratio.
Fig. 5. Beta coefcient plot from the PLSR (a), Residual plot for whole variables, PLS-VIP and PLS-SR method (b). PLSR, partial least squares regression; VIP, variable importance in projection; SR, selectivity ratio.
Table 1. Prediction results by partial least squares regression (PLSR), variable importance in projection (VIP), and selectivity ratio (SR) variable-selection methods for detecting pure and benzene-adulterated edible oils.