Forecasting Bank Failures and Stress Testing: A Machine Learning Approach

31 Pages Posted: 21 Sep 2017

See all articles by Periklis Gogas

Periklis Gogas

Democritus University of Thrace - Department of Economics

Theophilos Papadimitriou

Department of Economics, Democritus University of Thrace

Anna Agrapetidou

Democritus University of Thrace - Department of Economics

Date Written: September 19, 2017

Abstract

In this paper, we present a forecasting model of bank failures based on machine-learning. The proposed methodology defines a linear decision boundary separating the solvent from the failed banks. This setup generates a novel alternative stress testing tool. Our sample of 1443 U.S. banks includes all 481 failed banks during the 2007-2013 period. The set of explanatory variables is selected using a two-step feature selection procedure. The selected variables were then fed to a to Support Vector Machines forecasting model, through a training-testing learning process. The model exhibits 99.22% overall forecasting accuracy and outperforms the well-established Ohlson’s score.

Keywords: machine learning, bank failures, stress testing, forecasting

JEL Classification: G33, G21

Suggested Citation

Gogas, Periklis and Papadimitriou, Theophilos and Agrapetidou, Anna, Forecasting Bank Failures and Stress Testing: A Machine Learning Approach (September 19, 2017). Available at SSRN: https://ssrn.com/abstract=3039358 or http://dx.doi.org/10.2139/ssrn.3039358

Periklis Gogas

Democritus University of Thrace - Department of Economics ( email )

Komotini, 69100
Greece

HOME PAGE: http://www.econ.duth.gr/personel/dep/gkogkas/index.en.shtml

Theophilos Papadimitriou

Department of Economics, Democritus University of Thrace ( email )

University Campus
Komotini, 69100
Greece

HOME PAGE: http://econ.duth.gr/author/papadimi/

Anna Agrapetidou (Contact Author)

Democritus University of Thrace - Department of Economics ( email )

Komotini
Greece

Register to save articles to
your library

Register

Paper statistics

Downloads
332
rank
88,529
Abstract Views
1,051
PlumX Metrics