Mg element can make the aluminum alloy to obtain better mechanical properties and form a corrosion-resistant spinel film on the surface of the alloy, so that the alloy has better corrosion resistance. Therefore, exploring a method that can quickly and accurately detect the content of magnesium in aluminum alloy quantitatively is of great significance. In this paper, first, the Mg element in 17 aluminum alloy samples is detected and analyzed based on laser-induced breakdown spectroscopy (LIBS) technology. Then, Nd:YAG laser is used as light source, and the partial least squares (PLS) and random forest (RF) models are respectively established, and the prediction performance of the models is analyzed. The experimental results show that for the same test set, the correlation coefficient (Rp2) of the PLS model is 0.6809, and the root mean square error (RMSE) is 1.2042; the Rp2of the RF model is 0.8571 and the RMSE_{is 1.0918. In order to improve the prediction accuracy of the random forest model, this experiment screened the input variables according to the importance of the variables. When the wavelength point with variable importance greater than 0.11 is selected, Rp2 of the RF model based on variable importance is 0.9461, and the RMSE is 0.9534. Compared with the prediction result of the RF model, Rp2 is increased by 10.38%, RMSEis reduced by 12.68%, and the modeling time is reduced by 91.67%.}