imbalanced-seismic-data-classification

• data.rar: The data sets were downloaded from the data mining competion: AAIA'16 Data Mining Challenge: Predicting Dangerous Seismic Events in Active Coal Mines
• seismic data preprocessing.ipynb
• dara_pre.py
• model.py
• feature combined.ipynb
• MLP.py: Multilayer Perceptron model
• CNN_series.py: Convolutional Neural Network as the classifer.
• FCN.py: Fully Convolutional Neural Network as the classifer.
• res_net.py: Residual Network as the classifer.
• LSTM_FCN.py： LSTM combined with FCN as the classfier.
• classifier_train.py: train and test the classifer with the processed and filed data sets. For rebalance the training samples, SMOTE(Synthetic Minority Over-sampling Technique) was applied and 10-fold cross validation was used during the training period. Change the "NET", such as MLP, CNN_series or FCN, you can use different models to do the imbalanced data classification. The messures are confusion matrix, ROC AUC, G-mean, F1 score and so on.
• classify_train_without_features.py: Same with classifier_train.py, but the input data samples without features.