With the credit card still being a most widely excepted credit product, credit card companies have an essential responsibility to protect their customers from risks; both external (e.g. fraud) but also internal (e.g. irresponsible credit usage).
This credit card company has historical transactions at its disposal, which contain valuable information on credit behaviour. Investigating each transaction individually is not feasible; therefore, business rules were used. However, capturing all risk pattern in business rules is difficult since user behaviour changes over time.
Using machines learning, we developed a model to uncover the most important patterns from historical transactions to make a pre-selection of high-risk clients. This pre-selection of clients can serve as a good basis for further investigation of a credit expert, as such saving time in time consuming process.
- Efficiency in a labor intensive process
- Self-learning model which can be re-trained easily
- Reveal complex transaction patterns