Εφαρμογή μεθόδων εξόρυξης γνώσης σε τραπεζικά δεδομένα ελληνικών επιχειρήσεων
Application of data mining methods in banking data of greek enterprises
Καντά, Ελισάβετ Φ.
KeywordsΕλληνικές τράπεζες ; Λογιστική παλινδρόμηση ; Τεχνητά νευρωνικά δίκτυα ; Στατιστικές μέθοδοι
The economic crisis that started in Europe in 2008 and lasts until today in Greece raises the urgent need to monitor the risk a financial institution takes when it gives a loan. A model that can predict timely and valid loans that will become inactive in the near future is a vital tool for any banking organization. Our aim is to describe and compare such models by applying them to data from a large Greek bank that concern business customers. More specifically, we have developed a classification model that predicts if an enterprise’s loan becomes non-serving at the end of the next six months. We have applied to our data the method of Logistic Regression, Artificial Neural Networks, and Classification Trees using either all variables or using only those that were identified as significant after proper pre-processing. Specifically, stepwise and sensitivity analysis was applied to identify the variables that play an important role in the result.