There are many types of hams and their origins are different. This article uses laser-induced breakdown spectroscopy (LIBS) combined with machine learning algorithms to carry out research on the identification of ham origins. The LIBS spectrum data of 16 ham slice samples(4 Rugao ham samples， 5 Jinhua ham samples， 7 Xuanwei ham samples) were collected in the experiment， using K-Nearest Neighbor (KNN)， Support Vector Machine (SVM) and Deep Neural Network (DNN) classifies the origin of ham samples， and studies the dimensionality reduction processing of the spectrum data of the ham samples using Principal Component Analysis (PCA)， and then combines the KNN and SVM algorithms to classify the samples and the speed of modeling And the impact of forecast accuracy. The research results show that: KNN and SVM combined with PCA， the modeling and analysis time is greatly reduced； the average accuracy of the four classification methods of KNN， PCA+KNN， SVM， and PCA+SVM are 70.53%， 73.50%， 79.53%， and 80.42%， when using KNN and SVM combined with PCA， the classification accuracy is improved slightly； Using DNN to classify the ham samples， the classification accuracy rate can reach 85.56%， compared with KNN and SVM， DNN has a higher classification accuracy rate for ham LIBS spectrum data. The above show that LIBS combined with machine learning algorithm is feasible to distinguish ham samples from different origins.