Abstract: Determination of the Chinese liquors from different geographic origins is benefit for controlling liquor quality and safeguarding the interests of consumers. In this study, mass spectra of 131 Chinese liquor samples from different geographic origins were collected by the headspace (HS) solid phase microextraction (SPME) mass spectrometry (MS), without pre-treatment or chromatographic separation. By combination of partial least squares discriminant analysis (PLS-DA) and stepwise linear discriminant analysis (SLDA) methods, 27 characteristic ions are finally selected and the prediction ability of the SLDA is 99.2%. And then a back-propagation (BP) neural network and a support vector machine (SVM) recognition model are built, whose prediction accuracy are up to 96.2% and 97.7%, respectively. The parameter optimization result of BP neural network is logsig, trainlm and eight neurons in hidden layer, while the parameter c, g is 2 and 0.125 in the SVM model. According to the optimization procedure and prediction accuracy, the SVM model is superior to the BP neural network.