文摘
Near-infrared spectroscopy (NIRS), combined with diverse feature selection techniques and multivariate calibration methods, has been used to develop robust and reliable reduced-spectrum regressionmodels based on a few NIR filter sensors for determining two key parameters for the characterizationof roasted coffees, which are extremely relevant from a quality assurance standpoint: roasting colorand caffeine content. The application of the stepwise orthogonalization of predictors (an "old" techniquerecently revisited, known by the acronym SELECT) provided notably improved regression modelsfor the two response variables modeled, with root-mean-square errors of the residuals in externalprediction (RMSEP) equal to 3.68 and 1.46% for roasting color and caffeine content of roasted coffeesamples, respectively. The improvement achieved by the application of the SELECT-OLS methodwas particularly remarkable when the very low complexities associated with the final models obtainedfor predicting both roasting color (only 9 selected wavelengths) and caffeine content (17 significantwavelengths) were taken into account. The simple and reliable calibration models proposed in thepresent study encourage the possibility of implementing them in online and routine applications topredict quality parameters of unknown coffee samples via their NIR spectra, thanks to the use of aNIR instrument equipped with a proper filter system, which would imply a considerable simplificationwith regard to the recording and interpretation of the spectra, as well as an important economic saving.