ServicesShow__hero__img

Bad Loan Prediction

Loan companies often have a hard time predicting which loans are good and which are bad. By having an AI model that learns from past loans, we are able to build a predictive model that can reduce NPL down to less than 10%.

Business Challenges

Financial institutes have a huge volume of applicants applying for loans. Most of the applicants do not have record with the National Credit Bureau, in other words, they are the unbanked. Hence, it is impossible to know their credit scores.

That doesn't mean that we should reject all of these applicants, because many of these unbanked citizens contribute to a massive amount of potential loans.

Solution

We make an AI model to learn from past applicants whether they were good or bad loans. The model performs out of sample with accuracy of 93%, which helped the loan company to reduce their NPL down to less than 10%, an NPL lower than most banks.

Deliverables

  • A machine learning model to predict loan application.
  • A dedicated form for the user to have enter applicant information. When submitted, it triggers a machine learning inference and outputs a credit score of the person.

Expected
Outcome

A reduced NPL and higher profits.

Have a question?

Drop us a line and we will get back to you