In this project, I trained models to predict in-hospital death. See this link for details. The best model was a gradient-boosted classifier with a mean cross validation score of 0.37-0.38, beating the sample (SAPS-I) and random-predictor benchmarks.
- Make sure
condais installed. - Create the local environment:
conda env create -f environment.yml. - Run
jupyter laband openicu-survival.ipynb.