//= Titles\title(); ?>
Challenge
- Credit Card company was observing high risk customers which were suddenly going default within a month and fleeing country after doing high risky transactions and using up all their limit.
- The risky behavior was observed across two sets of customers- Intentional Low Vintage defaults where customer takes up the card to build up the charges and then leave the country or High Vintage customers which are stuck by hardship.
- The current early warning models were not attuned to the evolving market trends and had challenges in correctly identifying these accounts
- We were also observing lot of false positives (accounts incorrectly tagged as high risk) which created operational challenges as well as adverse customer impact.
Solution
- Built decision forest binary classifier model which identifies customer’s to be “high-risk” by analyzing their recent performance.
- Model is based on attributes from demographics, customer transactions, early pre authorized transactions, behavioral pattern account history and payment history. Did advanced feature engineering to create attributes based on business insights and data trends to better capture risky behavior indicators
- Model is able to capture non-linear relationships in the data as well as look at both short term and long term trends in behavior.
Result
- Reduced losses with increased efficiency in identifying “high risk” accounts.
- Enables prioritization among accounts by leveraging the model prediction score
- Assists collection efforts as it reduces the number of “false cases” and hence prioritizing resources for the “high risk” accounts