In order to effectively use available historical observation data for precipitation forecasting in the case of an uncertain cause of precipitation, a precipitation forecasting method was developed based on the naive Bayes algorithm. Using the Dongjiang Basin as an example, a rich set of features was constructed based on the basin’s precipitation data and meteorological knowledge. The forecasting accuracy of the proposed method was compared with those of the traditional time series method and the BP neural network method. The result shows that the proposed method outperformed both the traditional time series method and the BP neural network method.