Laser-induced breakdown spectroscopy (LIBS) can potentially be employed for remote-sensing in situ detection, which is a crucial technique for deep-space exploration for identifying the composition and content of material elements. The exploration of element composition and mineral distribution characteristics on the surface of Mars is the premise of studying the geological evolution and genesis of Mars. Before launching the Tianwen-1 mission, Mars-simulated exploration experiments were conducted on 15 categories of mineral samples using the Mars surface composition detector (MarSCoDe), and 1920 spectral datasets were collected. This study adopted an efficient classification model to verify the detection performance of the instrument using the artificial fish swarm algorithm (AFSA)-optimized support vector machine (SVM) (AFSA-SVM), which classifies 32 minerals, including igneous and sedimentary rocks and metal minerals. First, the principal component analysis (PCA) was adopted to reduce the dimension of the original spectral data, and the data was trained in AFSA-SVM. Second, AFSA optimized the parameters of SVM and achieved 99.56% mineral recognition accuracy. Finally, AFSA-SVM was compared with other algorithms, including the random forest (RF) algorithm, backpropagation artificial neural network (BPANN), and K proximity (KNN) algorithm. Their accuracy values are 95.60%, 95.80%, and 90.17%, respectively. The results show that the AFSA-SVM algorithm has advantages in assisting LIBS in identifying mineral targets.