Firm wanted to develop scorecard based on Bureau data to understand default risk of the customer who are applying for two-wheeler loan.
- What should be the default risk definition for the defaulters?
- What should be the most appropriate time period for model development.
- What are the final variables that need to be considered while model development?
- Logistic regression was used to make model.
- To increase the discrimination power of the model, the population was divided into two segments, so two models were prepared.
- One was thin segment as it had only those records which had one trade and another segment had more than one trade.
- 70% of the two-wheeler population was taken as model development and 30% of the two-wheeler population was taken as testing purpose.
Provided client with the scorecard model as well as data to enable them to take an informed decisions.