This is Team CVP's solution to NIH NCAT's Bias Detection Tools in HealthCare Challenge
This is an example of how you may give instructions on setting up your project locally. To get a local copy up and running follow these simple example steps.
- Python 3.8
- Requirments are in
scripts\requirements.txt
The script is using the following directory tree structure:
├── scripts/ # This is where `measure_disparity.py` and `mitigate_disparity.py` are
├── reports/ # Location for the html report generated by the script
├── input_model/ # Insert input model data here
├── output_model/ # This is where the mitigated model prediction is saved
├── data/ # Sample data to run the model
├── js/ # javascript for the team submission landing page
├── css/ # css files for the team submission landing page
├── assets/ # images for the team submission landing page
- Clone this repo
- Save a copy of the input in
input_model/
asinput_model.csv
.input_model.csv
should include one row per individuals with columns below:
* Model prediction (as a probability)
* Binary outcome (i.e. 0 or 1, where 1 indicates the favorable outcome for the individual being scored)
* Model label
* Sample weights
- Execute the python file
python scripts\measure_disparity.py
- An html report will be generated in
reports\measure_report.html
- If the report shows bias, users can mitigate the bias by running
python scripts\mitigate_disparity.py
- Html reports will be generated in
reports\mitigate_report.html
andreports\imbalance_report.html
Distributed under the BSD 3 License. See LICENSE.txt
for more information.
- Manpreet Khural - manpreetkhural@cvpcorp.com
- Cal Zemelman - calzemelman@cvpcorp.com
- Lauren Winstead - laurenwinstead@cvpcorp.com
- Wei Chien - weichien@cvpcorp.com
- Rose Anderson - roseanderson@cvpcorp.com
Project Link: https://github.com/cvp-bias-detection-healthcare/