Owing to the serious interference of soil moisture content in the detection techniques such as X-ray fluorescence spectroscopy (XRF) method, a support vector regression (SVR) correction prediction model is proposed based on the grey wolf optimization (GWO) algorithm. Subsequent to the preprocess of spectral data, a quantitative analysis model for determining the relationship among net peak area, moisture content, and cadmium content is established based on GWO-SVR. The GWO-SVR model is compared with other models. The results show that the SVR nonlinear model has a better decision coefficient and smaller errors than the linear regression model. Moreover, under the GWO optimization, each model index is improved. Compared with other optimization algorithms, GWO-SVR has less iterations, better fitting effect, and smaller prediction errors. The proposed model can provide an effective reference for the prediction of other heavy metals in soils and the correction of moisture content.