ICP Prediction from Routine Monitoring Data
Setting up your environment
Use python -m pip install requirements.txt to install the required packages.
How to run
Preprocessing
Preprocess your MIMIC data to include all relevant features in long format. Then place it into the datafolder as Datenbank_Werte.csv. Run the preprocess_data.ipynb notebook to preprocess all data.
How to train models
Run the training notebook train.ipynb. Modify parameters in the first cell blocks as you see fit. This notebook also includes extensive evaluation of the trained model and generates most of the plots.
How to tune hyperparameters
Run the tuning script, e.g.:
python tune.py model_type="rnn"
Once models have been tuned, check the analyze_tunings.ipynbjupyter notebook to generate plots of your tuning results.