LoanSmart

LoanSmart uses machine learning to assess loan eligibility and recommend the optimal amount based on the customer's history, empowering financial institutes to make smarter, data-driven decisions.

Comment

Inspiration

The inspiration for LoanSmart stemmed from personally witnessing people getting rejected for loans simply because they did not have a credit card or were biased against because of their features. This inspired us to create a project that could eliminate bias and make smarter data-driven decisions. With a surge in digital transactions, analyzing customers' financial behavior can help loan distribution and reduce risk.

What it does

LoanSmart is a data-driven machine learning system that evaluates a customer's loan eligibility and predicts a suitable loan amount. By analyzing the customer's past transaction pattern, pending balances, and other aspects of their financial history, the model helps lenders assess the risk levels that come with loaning money to each customer and make informed loan decisions.

How we built it

We collected and cleaned customer transaction data from Capital One's Hackathon API, Nessie, engineered some features, and trained two machine learning models:

  • A classification to predict loan eligibility.
  • A regression model to recommend the optimal loan amount. We then evaluated and optimized the models for optimal performance.

Challenges we ran into

Initially, we ran into the problem of not having a proper data source to train our model on. We later decided to make do with Nessie's Enterprise data, which contained several inconsistencies and missing values. That required a lot of extensive cleaning and preprocessing to get it suitable for training the model. There was also the difficulty with choosing the features that we choose to use that will be of the most help to training the model.

Accomplishments that we're proud of

  • Successfully compile a dataset that is usable and useful for the model
  • Implementing simple yet robust models with a high accuracy that will help serve the marginalized community

What we learned

  • The importance of a good and comprehensive dataset that is related to the issues at hand
  • The features that were important to loan eligibility and how easily bias can affect it
  • How to interact and effectively utilize APIs.

What's next for LoanSmart

  • Expanding the feature set to include more financial data such as credit scores, investments, and etc.
  • Developing a blockchain learning game that can help users learn more about financial literacy and investing
Share this project:

Updates