In order to identify chili peppers from different regions， a classification model for chili peppers based on a combination of untargeted metabolomics and machine learning was established. A total of 108 samples of Sichuan， Hunan， and Hebei peppers were selected as the research objects. The orthogonal partial least squares (OPLS-DA) method was used to analyze the differences in peppers from different regions， identifying 14 differential metabolites such as sugars， alcohols， organic acids， and amino acids. The differential metabolites were enriched and analyzed through the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway， revealing the citric acid cycle， glyoxylate， and dicarboxylic acid metabolism. The three metabolic pathways， including aminoacyl tRNA biosynthesis， differ significantly in chili peppers from different regions. 14 differential metabolites were screened using Lasso regression， and 8 important differential metabolites were identified. Two classification models， support vector machine (SVM) and random forest (RF)， were established， with an accuracy of 97% in the test set. The reliability and effectiveness of the random forest model and support vector machine model were validated using receiver operating characteristic (ROC) curves. This study provides a theoretical basis for identifying the origin of chili peppers and developing chili related products.