FAT Forensics Dashboard Example

⚠️ The dashboard has been migrated from Heroku to Render due to the former platform abandoning the free tier.

This repository holds the source code for a simple FAT Forensics dashboard built with dash. You can preview the deployment at https://fatf.onrender.com/; please allow 30 seconds for the dashboard to load. This example is based on a logistic regression model trained with scikit-learn on the Adult data set. For more information see the FAT Forensics: A Python Toolbox for Algorithmic Fairness, Accountability and Transparency paper published in the Software Impacts journal.

Preparing Data and Models

A pre-processed data set (together with the labels) and a trained model are included in the _data_model directory. These can be regenerated with the prepare_data_model.ipynb Jupyter Notebook included therein.

Local Deployment

The dashboard can be deployed locally by installing the dependencies

pip install -r requirements.txt

launching gunicorn

gunicorn --workers=2 app:server

and navigating to the local deployment accessible at http://127.0.0.1:8000/.