Mine disaster event detection method based on convolutional neural network

The invention relates to a mine disaster event detection method based on a convolutional neural network, which is suitable for coal mine disaster event detection. The method comprises two parts of model design and model training. A model design part mainly comprises a mixed feature input layer and a high-speed iterative hole convolutional neural network design part. A text word level, a characterlevel and an entity feature vector are input in a mixed manner to create a feature input layer with higher fine granularity and semantics and fully mine text deep semantics and structure information..An iterative method is adopted to stack DCNN to construct a deep model; the global feature vector is obtained, the model training efficiency is improved, the model rear-end framework selects the HighWay network to be connected with the Softmax classification layer to optimize the feature vector, the problems that the training deep network model is prone to over-fitting and difficult to convergeare solved, and the model de