Credit card companies are interested in offering their customers loans provided that the customer has the ability to return the loans.
When a credit card company wants to offer its customers loans, it is only interested in addressing customers who have a high ability to return the loans and a low probability to default.
We reviewed the internal and external (BDI an external credit score company) data sources. During the review process, we identified properties that could predict the customer’s ability to return the monthly payments.
We built a model that identified for each customer the probability of their inability to meet the monthly return payments of the loan. The process of building the model included a test of a number of statistical algorithms. The algorithm with the most accurate prediction ability was chosen. The most accurate algorithm was selected after comparing all the algorithms statistical results with statistical tools.
The final model projected each customer’s probability not to meet the loans payments. We defined a probability threshold over which the customers would not receive a loan. The maximum amount to be loaned was recommended by the model as well.