Laser induced breakdown spectroscopy (LIBS) is used to analyze the prepared fly ash samples, and support vector machine regression (SVR) model is used to predict the carbon content of fly ash. The structure parameters of radial basis function (RBF) kernel function and polynomial function are optimized by grid search method, and then SVR models based on internal standard element characteristic spectrum, full spectrum, and main element characteristic spectrum are established respectively. The research shows that SVR model of RBF and polynomial kernel function can achieve the same analysis accuracy under ideal structural parameters, but RBF can complete the model optimization quickly and is not easy to underfit. The analysis accuracy of the SVR model based on the characteristic spectrum of internal standard elements is similar to that of the internal standard method, and the SVR model based on full spectrum shows obvious overfitting phenomenon. The regression coefficient of the SVR model based on the characteristic spectrum of the main elements is 0.986, the root mean square error of correction is 1.79%, and the root mean square error of prediction is 2.57%, indicating that the model can effectively avoid underfitting and overfitting.