Credit Card Fraud Detection

Inspiration

Everyone depends on economic stability in some form to learn.

What it does

Our microservice will classify a credit card transaction as either fraudulent or legitimate.

How we built it

We implemented a simplistic linear regression algorithm. We took a minimalistic approach to the problem to maximize our efficiency over our limited time, focusing on picking the proper combinations of loss functions, activation and dropout over preprocessing our data.

Challenges we ran into

Our initial training resulted in a 31% f1-score, which was extremely concerning. Later down the road, we ran into over-fitting issues, which were addressed with an architecture change and sufficient dropout.

Accomplishments that we're proud of

We're proud of our performance despite any data augmentation or preprocessing.

What we learned

We learned that we can probabilistically categorize credit card transactions with a fair amount of certainty.

What's next for Credit Card Fraud Detection

I would like to add data preprocessing - augmenting our dataset, downsampling (our dataset is incredibly unbalanced), and experiment with other techniques like gradient-boosting.