Social justice is all about equality, whether it's about different sex, or different races. A lot of big tech companies have begun their journey of data analysis given the data and human resources they possess to identify the bias and inequality among the U.S. system. In justice system, inequality is more important than ever.
What it does
Our goal is to inform people of the racial bias in the U.S. Justice system and give the offenders a chance to know their score level beforehand.
How I built it
I used Jupiter notebook to build the analysis and test my models with scikit-learn and statsmodels. And then I used Python3 and Streamlit to write my web app. I used GitHub and Docker to make a container image and then deploy it onto Azure.
Challenges I ran into
Dataset I got is too small and don't have many variables about the offenders' demographics information. Therefore, I cannot get a high enough prediction accuracy score out of it. The models need improving a lot. The web app deployed on Azure is sometimes unstable because of some of the bugs from Streamlit. If you cannot connect it unfortunately, please take a look at another link here.
Accomplishments that I'm proud of
This is my first time building web apps and first time building a docker image and deploying it to the cloud platform. I searched for the dataset and idea for a month but finished the whole project almost in a week on my own.
What I learned
Racial bias is really a serious challenge to identify and making a system equally to everyone is difficult. Above automation, we need real people behind the score system to ensure the equality. A long way ahead of us to the great justice.
What's next for Racial Bias and Score Prediction of COMPAS Score
A better dataset and more informative prediction model with a higher accuracy score.