Credit risk is defined as the risk that borrowers will fail to pay its loan obligations. In recent years, a large number of banks have developed sophisticated systems and models to help bankers in quantifying, aggregating and managing risk. The outputs of these models also play increasingly important roles in banks’ risk management and performance measurement processes. In this study we try to tackle the question of default prediction of short term loans for a Tunisian commercial bank. We use a database of 924 credit records of Tunisian firms granted by a Tunisian commercial bank from 2003 to 2006. The K-Nearest Neighbor classifier algorithm was conducted and the results indicate that the best information set is relating to accrual and
cash![](http://cdncache-a.akamaihd.net/items/it/img/arrow-10x10.png)
-flow and the good classification rate is in order of 88.63 % (for k=3). A curve ROC is plotted to assess the performance of the model. The result shows that the AUC (Area Under Curve) criterion is in order of 87.4% (for the first model), 95% (third model) and 95.6% for the best model with
cash![](http://cdncache-a.akamaihd.net/items/it/img/arrow-10x10.png)
flow information.