Solvent-based mid-infrared spectroscopy paired with modern machine learning approaches for saffron authentication

This study demonstrates the use of solvent mid-infrared (MIR) spectroscopy with a MIRA analyzer, combined with modern machine learning/chemometric techniques for saffron quality and authenticity control. A total of 111 authentic saffron samples from six distinct geographical regions in Iran were analyzed. Saffron metabolite extraction was conducted using a modified solvent extraction protocol based on ISO 3632 standard, screened and optimized through design of experiments (DOE). The study focused on two main objectives: geographical origin discrimination and adulteration detection. For exploratory data analysis, principal component analysis (PCA) was applied. Partial least squares-discriminant analysis (PLS-DA) was used for geographical origin discrimination, achieving a 93.0 % accuracy in the test set, with the first derivative as one important preprocessing step. However, a higher accuracy of 95.5 % was obtained using random subspace discriminant ensemble (RSDE) without the need for preprocessing the MIR data. In the next phase, the detection of four common plant-based saffron adulterants, including safflower, marigold, rubia and even saffron style was carried out using the MIRA analyzer. Data-driven soft independent modeling of class analogy (DD-SIMCA) successfully differentiated between authentic and adulterated samples, achieving 100 % sensitivity and specificity. PLS-DA and RSDE were then employed to identify the type and level of adulterants, with RSDE clearly outperforming PLS-DA, achieving accuracy above 94.0 %, as compared to PLS-DA's accuracy of over 90.0 %. In conclusion, the combination of solvent-based MIR spectroscopy and modern chemometric techniques shows great potential as a reliable tool for saffron quality control at the point of need.