Comparative analysis of visible and near-infrared (Vis-NIR) spectroscopy and prediction of moisture ratio using machine learning algorithms for jujube dried under different conditions

• A rapid and non-destructive method is needed to evaluate the quality of dried fruits. • Vis-NIR spectroscopy has a high potential to evaluate fruits in different drying conditions. • Prediction of moisture ratio was performed with the use of machine learning algorithms. • A spectral reflectance technique was applied to the vegetation indices. • The best results were obtained in drying with the microwave method. Dried fruits are considered healthy because they are high in fiber and carbohydrates and contain low fat. In this study, jujube slices were dried using three different methods (open sun, closed shade, and microwave). Then, characteristics, such as color, spectral reflectance, vegetation indices (VIs), rehydration rate (RR), drying kinetics, moisture ratio (MR), and moisture content (MC) were measured and compared after using the above-mentioned drying methods. Also, the MR was predicted by the MC, and the drying rate (DR), drying times, and final thickness were predicted using the multi-layer perceptron (MLP), gaussian process (GP), k-nearest neighbors (KNN), random forest (RF), and support vector regression (SVR) algorithms. The drying times for jujube slices dried through the open sun closed shade, and microwave methods were 1680, 1140, and 24 min, respectively. Among the six mathematical thin-layer models, the Jena & Das model obtained a high coefficient of determination (R 2 ) with experimental data in the three drying methods (open sun, closed shade, and microwave). The results showed that closed shade drying had the highest RR (4.58), followed by open sun drying (4.40) and microwave drying (3.33). The microwave drying technique produced the lowest color change, with a value of 10.67. Shade drying resulted in the least light reflection across the spectrum, suggesting that the color of the product darkened as drying time increased. For MR prediction, the highest R-values belonged to MLP and RF, with values of 0.9997 and 0.9968, respectively. Regarding the attributes under study, microwave drying is a promising method for drying the jujube slices.