Discrimination between lard and other edibles fats is a challenging task for halal determination especially after the fats were heated at high temperature for a long period. In this study, three multivariate regression models such as partial least square regression (PLSR), principal component regression (PCR) and support vector machine regression (SVMR) were applied to evaluate the spectral data of FTIR (n=195) obtained from lard, chicken, beef, mutton and vegetable fats after heated at different conditions (120-240°C and 0.5-3 hrs). The regression of the Y-binary matrix was used to discriminate lard (as 1) and the others edibles fats (as 0). Kennard Stone (KS) algorithm selected a subset of the training set (n=145) and test set (n=50). The test set was used to validate the prediction ability of the suggested models. The obtained results showed the ability of the three proposed models to discriminate the heated lard simultaneously. The values of the R2 , adjusted R2 , root-mean-square error (RMSE) and root mean-square error of validation (RMSEV) showed a good results under Basic ATR correction transformation as PLSR (0.984, 0.977, 0.052 and 0.062); PCR (0.974, 0.971, 0.067 and 0.070), and SVMR (0.971, 0.959, 0.087 and 0.102) respectively. However, when using mean square error (MSE), it gives lower prediction error for PLSR (0.006), PCR (0.007) and SVMR (0.015). The results showed that PLSR as the best model for discrimination spectral data of lard and other edible fats after heating treatments for halal determination.

Discrimination of fatty acids (FAs) of lard in used cooking oil is important in halal determination. The aim of this study was to find the information related to the changes FAs of lard when frying in cooking oil. Quantitative analysis of FAs composition extracted from a series of experiments which involving frying cooking oil spiked with lard at three different parameters; concentration of spiked lard, heating temperatures and period of frying. The samples were analyzed using Gas Chromatography (GC) and Principal Components Analysis (PCA) technique. Multivariate data from chromatograms of FAs were standardized and computed using Unscrambler X10 into covariance matrix and eigenvectors correspond to Principal Components (PCs). Results have shown that the first and second PCs contribute to the FAs mapping which can be visualized by scores and loading plots to discriminate FAs of lard in used cooking oil.