Our experiements are conducted with MIMIC-IV 1.0 database. To run the code, you need to
- download the database;
- follow this tutorial to load MIMIC-IV into PostgreSQL database;
- run notebooks/generate_all_mimic_dataset.ipynb to read and save relevant data from the database;
- run src/data_prep.py to generate temporally-aligned 48H data samples in raw and anomaly-encoded format.
- Run train_cnn.sh to train CNN models
- Run train_xgb.sh to train RandomForest or XGBoost models