Predicting Failures of Russian Banks Using Parametric and Non-parametric Techniques
In this paper I compare three methods to predict bank failures in the Russian banking sector. Based on data ranging from 1997 to 2004 I test the predictive performance of Random Forests and Rotation Forests and compare it to logistic regression using four different time horizons for failure. The sample size ranges between 1,960 and 10,500 and includes 9 different financial ratios as predictors. I conclude that Random Forests outperform both Rotation Forests and logistic regression. Rotation Forests slightly outperform logistic regression in smaller failure time horizons. Overall, it can be concluded that all three models perform well in comparison to similar models in the literature.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted under the conditions of the Creative Commons Attribution-Share Alike (CC BY-SA) license and that copies bear this notice and the full citation on the first page.