Applicant-Insight

• 

  This is what the tool looks like for the user

Inspiration

Most credit tools are built for banks, not applicants. We wanted to make something that gives regular users a better idea of what kind of applicant profile they resemble without turning it into another approval prediction system.

What it does

Applicant Insight groups similar credit applicants into clusters using machine learning. Users enter anonymized applicant information and the app shows which applicant group they most closely match.

How we built it

We used Python, pandas, scikit-learn, and Streamlit. The data was cleaned, scaled, and one-hot encoded before training K-Means clustering models. We also used PCA to help visualize the clusters.

Challenges we ran into

The biggest challenge was working with anonymized data. Since the feature meanings were hidden, it made preprocessing, fairness decisions, and cluster interpretation a lot harder than usual.

Accomplishments that we're proud of

We’re proud that we built a full working pipeline from preprocessing and clustering to a usable frontend in a short amount of time. We also put a lot of thought into fairness and feature selection instead of just throwing a model together.

What we learned

We learned how important preprocessing is for clustering and how difficult fairness decisions can become when you don’t fully know what the features represent. I learned a lot about K-means and what K really means. lol

What's next for Applicant-Insight

We want to improve cluster explanations, add better visualizations, and experiment with other clustering methods to see if we can find even more meaningful applicant patterns. Possibly find more data.

Built With

• numpy
• pandas
• scikit-learn
• seaborn
• streamlit