In order to fast discriminate adulterated Fenyangwang wine, taking Fenyangwang wine with 53%vol of alcohol concentration as example, BP neural network and least squares support vector machine (LS—SVM) discriminant analysis models were established based on near infrared spectroscopy combined with chemometric methods. The spectra of 180 and 120 adulterated Fenyangwang wine samples were collected. Savitzky—Golay (SG) convolution smoothing was used to pre-treat the spectral data. Seven and eleven principal component factors were extracted respectively by using principal component analysis (PCA). Then, BP neural network and LS—SVM were used in discriminant analysis of unknown samples. Results showed that the accuracy of SG—PCA—BP neural network was up to 100%. The accuracy values of SG—PCA—LS—SVM models for two experimental groups were 84.4% and 83.3%, respectively.