3.5. Quantitative determination of sesame oil in the mixture
According to the detection of adulterated sesame oil, the other
important problem would be focus on quantitative determination
of sesame oil regardless of the type and amount of adulterant.
Therefore, the calibration model obtained for sesame oil in this
study requires more attention. The modeling and prediction of
mixture composition is done by partial least squares (PLS) regression,
a technique widely used in the chemometric literature for
multivariate analysis. The first step in the PLS determination of
the sesame oil in the mixture involves constructing the calibration
matrix. In this work, the normalized percentages of ten fatty acids
measured by gas chromatography are used for constructing calibration
matrix. 25% of 740 adulterated sesame oil samples are randomly
selected as validation set including 28 mixtures of rapeseed
oil and sesame oil, 46 mixtures of soybean oil and sesame oil, 37
mixtures of sunflower seed oil and sesame oil, 37 mixtures of cotton
oil and sesame oil and 37 mixture of palm oil and sesame oil,
respectively. To select the optimal number of factors for the PLS
models, the leave-one-out cross-validation method is employed.
The prediction error of sesame oil in the validation set is calculated
as the root-mean-square error (RMSE) of the prediction concentration
and can be described as