Twelve types of ores were identified using laser-induced breakdown spectroscopy combined with the principal component analysis-particle swarm optimization-support vector machine (PCA-PSO-SVM) algorithm. A Savitzky-Golay filter was used to smooth the spectrum, and the segmented eigenvalue extraction method was used to perform baseline correction on the spectrum. The first 25 principal components reduced by PCA were selected as the input to the PSO-SVM classification model, and the best recognition accuracy rate for the 12 types of ore was 100%. The PCA-PSO-SVM model was compared with two classification models, i.e., principal component-linear discriminant analysis and a PCA-particle swarm optimization-error back propagation neural network. Experimental results showed that the recognition accuracy of the PCA-PSO-SVM classification model was the highest with an average recognition accuracy rate of up to 99.90%.