Efficient and accurate risk prediction for credit operations

This use case is provided by Banco de Crédito Social Cooperativo S.A. (BCC), a Spanish rural savings bank and credit union, and will deal mainly with the measurement of clients’ probability of default. In the current context of the financial crisis, to be able to identify risky clients accurately is of vital importance for a financial institution; having a proper knowledge of clients makes it possible to take preventive actions and avoid future losses.

BCC maintains databases about all the customers serviced by the institution. These databases include demographic data, records of activity in accounts, data about credit operations (loans, credit cards, credit lines, etc.), clients’ payment history, etc. There are about  1000 variables for 4m clients. This is a massive amount of information that is not possible to treat with current techniques. Nowadays, only part of the available variables are considered and  no dependence structure between them is taken into account. A group of analyst identify the most promising factors that influence the future payment of a  credit obligation and construct statistical models with them. The technique employed is usually logistic regression, which  is the state-of-the-art method in the sector.

The AMIDST impact on future risk prediction for credit operations

BCC plans to use the AMIDST tool to asses risk for all the customers reflected in the database, making use of all the variables and taking into consideration the possible dependences between them. This will be possible with the use of more advanced techniques as hybrid Bayesian networks. Extracting all possible knowledge about our clients from the massive amounts of information that are recorded in our IT systems could bring BCC an important competitive advantage. It is expected that, if AMIDST goals are achieved, losses due to defaulting clients could be reduced by 7.5m € every year.

The AMIDST solution will allow us to update clients' risk prediction with a much more efficient frequency, which is important because clients’ information is constantly changing.  Also, it will be possible to update the credit risk model in a faster way. Efficiency in these tasks is expected to have an important improvement. Not only will the AMIDST tool be able to identify risky clients more accurately, but also to extract profiles of the clients with less credit risk. This is also very valuable information because BCC could then target these clients in its marketing campaigns, potentially obtaining better results and higher benefits on performance.

To sum up, the AMIDST research project is an important part of BCC’s continued commitment to innovation and best practices in risk management.

This website uses cookies. Cookies are only used for handling login and identifying registrered users, not for visitor tracking.