Sepsis Framework

A framework for dealing with the physionet 2019 sepsis dataset and making predictions to be scored against the pre-defined utility function.

My suggestion is to setup a virtual environment folder at /env/ in the root directory of this project. Install the requirements with

pip install -r requirements.txt

Create a data/ folder in the project directory, make this a symlink if you wish to store it somewhere else, then create a data/raw folder.

1. Run src/data/download.py to download the raw data into data/raw.
2. Run src/data/make_frame.py to convert the data into a dataframe format useful for visualisation in notebooks.
3. Run src/data/preprocess.py to perform various preprocessing steps (that one may want to change if running your own models) that includes simple feature derivation, and converts the ragged data into a nan filled tensor.
4. Run src/models/predict_model.py to run a simple model and get cross-validated scores.