Abstract: Rapid analysis of the causes of water outburst and accurate determination of its source are key to the management of water outburst disasters.By applying deep neural network theory and combining water chemistry analysis data from mine aquifers, eight feature ions were selected as discriminating factors for the identification of water sources in the mine, forming the input state space of the deep neural network, and water source categories were selected as sublabels, forming an output state space of a deep neural net, thus constructing a deep network model for outburst water sources identification.Trained on a learning sample of 2 734 samples of water sources, the model achieved recognition accuracy of over 95% on the training set.The method is applied to identify the water source in Huainan mining area.The results show that the depth neural network model has good classification performance and high prediction precision.