Soil moisture is one of the important factors affecting the growth of crops, accurate prediction of soil moisture in the critical period of crop growth is an important part of the field management. In this study 15 prediction factors were selected from meteorology, topography and soil properties from March to May in the winter wheat growing area from 2014 to 2016 in Baoji of Shaanxi Province, soil moistures in 0–20 cm and 20–40 cm soil layers were predicated and compared by using the established PCA-SVR (Principal Component Analysis-Support Vector Regression) model and Random Forest (RF) regression model. The results showed that prediction accuracies in 0–20cm and 20–40cm soil layers were 92.899% and 92.656% for PCA-SVR model, 87.632% and 87.842% for RF regression model, with the corresponding RMSEs of 7.521 and 8.011 for PCA-SVR model, 10.759 and 11.042 for RF regression model, respectively, indicating that PCA-SVR model had better predictive ability on soil moisture of winter wheat in Baoji, particularly for 0–20 cm soil layer.