Turmeric is an indispensable culinary spice in different cultures and a principal component in traditional remedies. Toxic metanil yellow (MY), acid orange 7 (AO) and lead chromate (LCM) are deliberately added to adulterate turmeric powder. This work compares the ability of multivariate chemometric models with those of artificial intelligent networks to enhance the selectivity of spectral data for the rapid assay of these three adulterants in turmeric powder. Using a custom experimental design, we provide a data-driven optimization for the sensitive parameters of the partial least squares model (PLS), artificial neural network (ANN) and genetic algorithm (GA). The optimized models are validated using sets of genuine turmeric samples from five different geographical regions spiked with standard adulterant concentrations. The optimized GA-PLS and GA-ANN models reduce the root mean square error of prediction by 18.4%, 31.1% and 55.3% and 25.0%, 69.9% and 88.4% for MY, AO and LCM, respectively.
This journal is © The Royal Society of Chemistry.