Customer Default Prediction

Key Highlights

  • Credit score is effected by a lot of factors and hence, a low score need not always mean a bad customer. Model driven approach helped the bank to tap these customers.
  • As the existing customers are also tracked, the chance of defaulting is minimized, and the loss is contained.
  • Similar approach was adopted for loans as well.

Challenges

The bank is relying solely on credit score for making the decision of offering credit card. Once offered, no process is in place to identify whether the customers can default or not.

Approach

It is a long established fact that a reader will be distracted by the readable content of a page when looking at its layout. The point of using Lorem Ipsum is that it has a more-or-less normal distribution of letters, as opposed to using ‘Content here, content here’, making it look like readable English. Many desktop publishing packages and web page editors now.

  • Fetch all the relevant details including salary, credit score, # credit cards, payment history etc. Enriched the data, did feature engineering with domain and stats driven approach
  • Build ML model and get a probability score of defaulting. If the probability is more than 0.4, then deny the card. Else issue the card even if the credit score is low.
  • A new model was built for predic>ng defaulters among the exis>ng customers based on the history of recent purchases, billing details etc. Track the card usage and payment frequency for the existing customers and identify the probability of defaulting. If the probability is high, reduce the credit limit.

Tech Stack

Python, Machine Learning, AWS cloud platform

Download 14+ Case Studies