Artificial neural network (ANN) is applied to make categorical forecasting of short-term precipitation in Meiyu period of Nanjing. At first, based on the weather characteristics of Meiyu period, in terms of the statistical and dynamic method, the predictors are searched in data of HLAFS. Secondly, utilizing import predictors that are selected by two methods trains the ANN, and then five days precipitation data are taken out in order to test the level forecast of precipitation. The results show that the simulating effect is very well and the ability of short-term precipitation forecast is obviously improved.