Objective To improve the classification accuracy of pork species by building an identification model based on near-infrared spectroscopy and PCA-DBN-SVM.Methods Combining the near-infrared spectroscopy characteristics of pork, principal component analysis (PCA) is used for dimensionality reduction and feature extraction, and DBN-SVM is then applied for classification and recognition to construct a pork species identification method that integrates near-infrared spectroscopy characteristics with PCA-DBN-SVM model.Results Compared with the KNN model, RF model, ELM, and DBN combination model, the PCA-DBN-SVM model has the highest classification accuracy of pork species, which reaches 99.91%.Conclusion The PCA-DBN-SVM model exhibits superior classification accuracy.