Abstract: Time series of hydrological forecasting data mining,based on grey theory and combined grey theory and neural network model.Original sequences transformed by logarithm-cube to meet the overlay need at first,then predicted by the model of grey forecasting model GM（1,1）.Due to the GM（1,1） belongs to linearity forecasting,introduces the neural networks combining with grey forecasting model to advance forecasting precision compare to the grey model.Experiments based on DuJiangyan Min-jiang River prove that the forecasting effect and rousting of the hybrid forecasting model is superior to traditional forecasting models.