Base code used for sepsis ML analysis
Python repository for sepsis machine learning model analysis (logistic, KNN, RF, XGB, SVM)
To run analysis:
- adjust file name in config.py and run
- run classifiers.py to generate functions for models
- adjust runXClassifiers.py to match desired variables, and output files (runClassifiers.py can be used for a non cross validated, non bagged version of the models). -to compare to clinical feature only model, adjust variables to just clinical features and result file names accordingly
- feature_importance.py can be used to generate feature importance rankings from the random forest model
Remaining files are used for data visualization
- Generate_Bar_Plots.py creates bar plots to compare results between two sets of models (eg clinical feature and full features)
- Generate_curves does the same for AUC-ROC curve, and can also plot AUC-ROC curves for multiple models together
- Generate_single_bar_plot is used to plot feature importance
- Generate_demographics generates a "Table 1", and generate_histogram generates a demographics histogram, although these are fully based on the original dataset for this project