Predicting Bank Financial Failures using Random Forest

Abstract

In 1994, there was an economic crisis in Turkey. Many banks were declared failed because of the negative impact from the crisis. The failure of individual banks has a huge impact on the real sector, households and can even cause knock-on effects for other banks. Therefore, it is important to predict bank failure. The 2009, Boyacioglu, Kara, and Baykan had predicted bank failures in the period 1994- 2004 using CAMELS as a predictor variable. In their research, they used Artificial Neural Network (ANN), Support Vector Machine (SVM) and multivariate statistical methods. However, in this research we will make novelty by using random forest as classifier method to predict bank failures in Turkey. Based on our results, random forest has 100\% accuracy in training performance, if we compare to other models that Boyaciaglu, Kara and Baykan (2009) paper, random forest is not inferior to other methods, if they must have to preparation data with normalization, in random forest we don’t need to do that. In testing set, random forest has 94\% accuracy with full ratio and 96\% accuracy with 6 ratios in its performance. Even random forest does not have accuracy more than other methods that have been used, but its accuracy is not far away with others. Then, random forest has property which is can measure important variable as predictor bank fimancial failure, in this research we get $\mathrm{C}\mathrm{A}_{2}, \mathrm{E}_{1}, \mathrm{C}\mathrm{A}_{3}, \mathrm{S}\mathrm{M}\mathrm{R}_{\mathrm{I}}$, SMR2, $\mathrm{E}_{2}, \mathrm{C}\mathrm{A}_{1}, \mathrm{L}_{1}, \mathrm{E}_{3}. \mathrm{A}\mathrm{Q}_{2}$ are six most important variables.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 25 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.