Garlic medicinal plant powder is significant from a commercial perspective and is widely used as an additive in the food and pharmaceutical industries. Spectroscopic techniques serve as non-destructive methods for assessing the quality of food products, medicinal plants, and related products. In the present study, the efficiency and potential of two widely used modes, reflectance and intractance, in near-infrared spectroscopy in the range of 936 to 1660 nm were compared to assess the feasibility of distinguishing fresh garlic powder. In each spectroscopy mode, 120 spectra in 2 replicates were obtained from garlic powder samples, resulting in a total of 240 spectra. Prior to modeling, 25% of the spectra were randomly selected for validation, while the remaining spectra were used to compile the models. To mitigate potential noise, the impact of common pre-processing methods on the performance of the artificial neural network (ANN), support vector machine (SVM), and knearest neighbors (KNN) classifiers was investigated. The principal components analysis (PCA) technique was employed to reduce the dimensionality of the spectral variables, and the first four principal components were used as classifiers inputs. In the intractance spectroscopy condition, the SVM and KNN classifiers separated spectra obtained from powders at 3 days, 3 months, and 12 months with 100% accuracy. The ANN classifier achieved 100% accuracy in distinguishing the mentioned spectra in all preprocessing conditions under investigation, except for the raw spectra (without preprocessing). Near-infrared spectroscopy in the 936-1660 nm range, combined with chemometrics, is effective for quickly detecting the freshness of garlic powder. Considering its ease of application in both industrial and laboratory settings, intractance spectroscopy mode is superior to reflectance mode.
