iPsRS project --> Link to the paper "A Fair Individualized Polysocial Risk Score for Identifying Increased Social Risk in Type 2 Diabetes"
Codebase for the iPsRS Fairness project.
- Describe any prerequisites, libraries, OS version, etc., needed before installing program.
- ex. Windows 10
- How/where to download your program
- Any modifications needed to be made to files/folders
./1.tuning_sheduler.sh
./2.parameter_checking_scheduler.sh
./3.bootstrapping_scheduler.sh
./4.assessment_scheduler.sh
./5.unfairness_mitigation_scheduler.sh
...
Modify the json files in the Settings to build your experiments.
Modify the {#.xx.sh} to run what you want to run.
Contributors names and contact info @Yu
-
0.1
- Initial Release
- Code Structure Reorganization
-
0.2
- Automated Pipeline, Tuning -> Assessment -> Mitigation
-
0.3
- Automated Pipeline, Tuning -> Checking -> Bootstraping -> Fairness Assessment -> Fairness Mitigation
- The framework is more flexiable, the user can DIY the base model (e.g., xgboost, catboost).
-
0.3.1
- Align model output path
- Reconstruct model.py
- Reconstruct Settings
-
0.4.0
- add feature importance analysis (SHAP)
- add Post analysis
- add causal analysis (double robust learning)
- The above mentioned functions are implemented in jupyter notebook.
TBD
Inspiration, code snippets, etc.