Figure 2.

Scheme of the analytical framework in NIR spectroscopy on the example of quantitative analysis. A qualitative approach (i.e., classification) is analogous; however, reference data might be arbitrary in that case (e.g., “authentic” vs. “non-authentic” product; “bad” vs. “good” quality product, etc.). The ultimate outcome is the prediction of the property of the “unknown sample” (i.e., for which the inefficient reference analysis was deliberately skipped) towards which the model was calibrated.