Abstract:The quality attributes of wine are closely related to its origin. In recent years，the issue of origin fraud has seriously disrupted the wine market.Therefore，it is necessary to establish a fast，accurate，and realtime method for identifying wine origins. In this study，ultraviolet-visible（UV-Vis）spectroscopy，near-infrared（NIR）spectroscopy，and chemometrics were combined to establish a rapid and accurate identification method for red wine origin. The spectroscopic information of 108 Cabernet Sauvignon red wine samples from three different regions，Huaizhuo Basin，eastern foothills of Helan Mountain in Ningxia，and Bohai Bay，was collected by UV-Vis spectrophotometer and micro-NIR spectrometer. Firstly，standard normal variable（SNV），first derivative（FD），second derivative（SD），and different methods were utilized to preprocess two kinds of spectral data. The discrimination models of random forest（RF），K-nearest neighbor（KNN），and least squares support vector machine（LS-SVM）were built. Comparing the classification effect of the models，the combination of optimal preprocessing and modeling methods was achieved. Secondly，based on the obtained optimal discrimination model，the competitive adaptive reweighted sampling（CARS）and uninformative variable elimination（UVE）were adopted to extract the feature bands of spectral data and build discrimination models. The results showed that the recognition rates of the correction set and the verification set of the SNV-FD-UVE-LSSVM model based on UV-Vis spectrum were 97.53% and 96.30%，respectively，and those of the FD-CARS-KNN model based on NIR spectra were 98.76% and 96.30%.The study showed that combining UV-Vis spectroscopy and NIR spectroscopy with machine learning had the potential to be a simple，fast，and low-cost tool for identifying the geographical origin of Chinese red wine.