Our team is taking a Machine Learning class this semester and we wanted to utilize everything we learned to an interesting and impactful real world issue.

What it does

Our project uses multiple models to predict the expectation of default on a loan. Then we use the multiple predictions to come to a final conclusion. So with 6 different models running at around 75-80% accuracy, we use the most common score and this gets our prediction accuracy up to 85%.

How we built it

Using Colab Jupiter notebooks and coding in python, with tensorflows libraries.

Challenges we ran into

We ran into some issues implementing our Neural Network and bringing all of the different models together.

Accomplishments that we're proud of

Having a working system that performs well.

What we learned

Our studies in the classroom have prepared us well to tackle real world problems

What's next for Leveraging Machine Learning to Predict Loan Defaults

Building a clean front end and monetizing the predicting model to help banks accurately assess risk in clients.

Built With

Share this project: