Neural Networks in Credit Risk Management

Maciej Kokorniak
PKO BP SA & Poznan University of Economics
Poznan, Poland


Neural Networks (NN) were elaborated to imitate the human brain competence for patterns classification, forecasting or making decision on a basis of a past experience. The main goal of my presentation is to demonstrate the usefulness of NN for Credit Risk Management Systems. I will describe my studies based on the real data stored in the books of the regional branch of the PKO BP SA (the largest Polish bank). The data collected during the nearly two year period (July 1, 2002 to March 31, 2004) allowed to conduct studies not only on through-the-door populations (which makes the basis for the most of publications on the subject) but also on the general population of credits. The results of my investigation suggest that NN may be competitive in the area of behavioral credit scoring systems over other data mining classifiers. I will show that the knowledge obtained thanks to my studies may be very useful in a current work of bank branches as well as in superior bank units.

Colloquia Series page.