COMPAS Recidivism & Bias

A neural network that predicts recidivism from ProPublica's COMPAS data while staying fair to minorities and vulnerable populations.

The objective of this project was to train a neural network on the ProPublica data (Correctional Offender Management Profiling for Alternative Sanctions) to predict recidivism — the likelihood of an offender committing a crime in the future — while avoiding human biases and remaining fair to minorities and more vulnerable populations.