Financial Distress Prediction: Zmijewski (1984) vs. Data Mining

Authors

ŠTĚRBA Martin ŠIŠKA Ladislav

Year of publication 2020
Type Article in Proceedings
Conference Proceedings of the International Scientific Conference of Business Economics Management and Marketing 2019
MU Faculty or unit

Faculty of Economics and Administration

Citation
Web https://webcentrum.muni.cz/media/3220002/sbornik-2019-105-converted.pdf
Keywords financial distress; data mining; neural networks; bankruptcy
Description The study re-estimates the Zmijewski's (1984) prediction model of financial distress with techniques offered by data miners. Namely logistic regression, neural network and decision tree models are applied to the training dataset consisting of approx. 130 thousand annual observations of financial ratios from non-financial companies residing in Czechia. Area under ROC curve (AUC) computed from similarly large independent testing set served as a measure of the predictive power of each alternative model. Our findings reveal the potential of neural networks to slightly, but statistically significantly increase the prediction power of the model. But this benefit goes in expense of complexity and lower interpretability of neural networks.
Related projects: