Elman Nets for Credit Risk Assessment
Nowadays the correct assessment of credit risk is of particular importance for all financial institutions to ensure their stability. Thus, Basel II Accord on Banking Supervision legislates the framework for credit risk assessment. Linear scoring models have been developed for this assessment, which are functions of systematic and idiosyncratic factors. Among statistical techniques that have been applied for factor and weight selection, Neural Networks (NN) have shown superior performance as they are able to learn non linear relationships among factors and they are more efficient in the presence of noisy or incorrect data. In particular, Recurrent Neural Networks (RNN) are useful when we have at hand historical series as they are able to grasp the data’s temporal dynamics. In this work, we describe an application of RNN to credit risk assessment. RNN (specifically, Elman networks) are compared with two former Neural Network systems, one with a standard feed-forward network, while the other with a special purpose architecture. The application is tested on real-world data, related to Italian small firms. We show that NN can be very successful in credit risk assessment if used jointly with a careful data analysis, pre-processing and training.
KeywordsCredit Risk Recurrent Neural Network Banking Supervision Credit Line Basel Committee
Unable to display preview. Download preview PDF.
- B. Back, T. Laitinen, K. Sere, and M. V. Wezel (1996). Choosing bankruptcy predictors using discriminant analysis, logit analysis, and genetic algorithms. In Proceedings of the first International Meeting on Artificial Intelligence in Accounting, Finance and Tax, pp. 337–356Google Scholar
- Basel Committee on Banking Supervision (2006). International convergence of capital measurement and capital standards a revised framework comprehensive version. Technical report, Bank for International SettlementsGoogle Scholar
- C. Bishop (2005). Neural Networks for Pattern Recognition. Oxford University PressGoogle Scholar
- A. Fan and M. Palaniswami (2000). A new approach to corporate loan default prediction from financial statements. Proceedings of Computational Finance Google Scholar
- A. Hamerle, T. Liebig, and D. Rösch (2003). Credit risk factor modeling and the basel II IRB approach. Banking and Financial Supervision. Deutsche Bundesbank. Discussion Paper, 2Google Scholar
- S. McCulloch and W. Pitts (1943). A logical calculus of the ideas immanent in nervous activity. Bullettin of Mathematical Biophysics, 7:115–133Google Scholar
- M. Mitchell (1996). An Introduction to Genetic Algorithms. MIT PressGoogle Scholar
- M. Odon and R. Sharda (1990). A neural network model for bankrupcy prediction. In Proceedings of the international joint conference on neural networks, pp. 163–168Google Scholar
- S.L. Pang, Y.M. Wang, and Y.H. Bai (2002). Credit scoring model based on neural network. In Proceedings of the First International Conference on Machine Learning and Cybernetics, pp. 1742–1746Google Scholar
- Li Rong-Zhou, Pang Su-Lin, and Xu Jian-Min (2002). Neural network credit-risk evaluation model based on back-propagation algorithm. In Proceedings of the First International Conference on Machine Learning and Cybernetics Google Scholar
- D.E. Rumelhart, G.E. Hinton, and R.J. Williams (1986). Learning internal representations by error propagation. In Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations, pp. 318–362. MIT Press, Cambridge, MA, USAGoogle Scholar
- Z. Yang, M. Platt, and H. Platt (1999). Probabilistic neural networks in bankrupcy prediction. Journal of Business Research, 44(2)Google Scholar