The Xin’anjiang model and Support Vector Machine(SVM)model were applied to the real-time flood forecasting of 4 basins in Zhejiang Province and Shaanxi Province. The K-Nearest Neighbor algorithm was used to correct the results of Xin’anjiang model. The forecasting results of different basins with two models were compared and analyzed by the evaluation indices of certainty coefficient, flood peak time error, flood peak relative error and mean square error. Further study was made to analyze the forecasting accuracy of two models in different forecast periods. The results show that Xin’anjiang model and Support Vector Machine model have their own advantages in the real-time flood forecasting of different basins. The accuracy of Support Vector Machine model is more susceptible to the accuracy of precipitation forecasting. Xin’anjiang model performs better in the case of long forecast period. With the forecast period reduced, the accuracy of Support Vector Machine was obviously improved. Meanwhile, Support Vector Machine model has a higher forecasting accuracy in the short forecast period. However, in semi-humid and semi-arid basins where the flood forecasting is difficult, the Xin’anjiang model and Support Vector Machine model have high accuracy in both calibration periods and real-time forecasting processes.