The linear regression (LR) and artificial neural network (ANN) methods are compared with each other in forecasting precipitation types under the same conditions. The selected predictors are surface air temperature and dew point when and 6 hours before precipitation happens, and the types include rain, sleet and snow. The observation data from 756 weather stations of the National Meteorological Centre, CMA during 2001-2011 are used, of which the data of 2001-2010 are used to test the methods and the 2011 data are used to verify the forecasting effects. The results show that both of the LR and ANN methods have prediction capacity for the three precipitation types of snow, rain and sleet. The predictability of snow is the best, then is rain, and the worst is sleet. Forecasts for the rain and snow separatrix forcasted by the two methods in the North of China are better than that in the South of China. The forecasting effect of ANN method is superior to that of LR method under the same conditions. When the temperature and dew point are forecasted correctly, the ANN method can be used to predict the rain and snow separatrix in the North of China exactly.