Abstract
We examine whether corporate bankruptcy predictions can be improved by utilizing the recurrent neural network (RNN) and long short-term memory (LSTM) algorithms, which can process sequential data. Employing the RNN and LSTM methodologies improves bankruptcy prediction performance relative to using other classification techniques, such as logistic regression, support vector machine, and random forest methods. Because performance indicators, such as sensitivity and specificity, differ depending on the methodology, selecting a model that suits the purpose of the bankruptcy predictions is necessary. Our ensemble model, a synthesis of all methodologies, exhibits the best forecasting performance. In the test sample for the ensemble model, none of the observations with a default probability of less than 10% defaults within one year.
Similar content being viewed by others
Availability of data and material
The data used in this study is downloaded from Wharton Research Data Services. Subscription required.
Code Availability
This research is conducted by installing scikit-learn v.0.23.1 and tensorflow v.2.3.1 on Python v.3.8.5. The code used in the study can be disclosed upon request.
Notes
Detailed explanations of these metrics are provided in the Appendix.
This sample period is the best option given our computational resources. However, a preliminary study analyzing the logistic regression, support vector machine, and random forest methodologies from January 1961 to December 2019 finds that the random forest model, which is an ensemble learning model, performs the best. That result is consistent with our current empirical results.
We use this simple method to fill in missing data because correcting missing values is not the main focus of this study. We choose not to drop missing variables because we would lose 16.12% of the total available observations if we did. Moreover, because we need 12 months of continuous observations for our analysis, the potential scope of data loss if we dropped all missing variables may be even larger.
Although these methodologies do not account for the dynamics of the features, they can still input the information. We try using a total of 96 explanatory variables in these models, but we nevertheless cannot construct forecasting models, mainly owing to memory overflow. The RNN and LSTM methods can leave only necessary information considering the order of the explanatory variables, whereas other methodologies need to estimate all relevant parameters. The authors are grateful to an anonymous referee for pointing out this issue.
We thank an anonymous for suggesting ways to structure the training data.
In Table 2, Logistic stands for logistic regression, SVM stands for support vector machine, RF stands for random forest, RNN stands for simple recurrent neural network, LSTM stands for long short-term memory, and Ensemble is a model that averages the predicted probabilities calculated by each individual model. Since Table 2, all figures and tables use the same notations for the models: Logistic (logistic regression), SVM (support vector machine), RF (random forest), RNN (simple recurrent neural network), LSTM (long short-term memory), and Ensemble (ensemble model).
Rather et al. (2015) also show that a hybrid model that combines RNNs and other linear models performs the best in predicting stock returns.
Because the training period of 2007–2014 includes the global financial crisis, these results are acceptable.
We appreciate an anonymous reviewer’s suggestions regarding this table’s layout.
References
Altman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance, 23(4), 589–609
Aretz, K., Florackis, C., & Kostakis, A. (2018). Do stock returns really decrease with default risk? New International Evidence. Management Science, 64(8), 3821–3842
Barboza, F., Kimura, H., & Altman, E. (2017). Machine learning models and bankruptcy prediction. Expert Systems with Applications, 83, 405–417
Beaver, W. H. (1966). Financial ratios as predictors of failure. Journal of Accounting Research, 4, 71–111
Bonfim, D. (2009). Credit risk drivers: Evaluating the contribution of firm level information and of macroeconomic dynamics. Journal of Banking & Finance, 33(2), 281–299
Brogaard, J., Li, D., & Xia, Y. (2017). Stock liquidity and default risk. Journal of Financial Economics, 124(3), 486–502
Campbell, J. Y., Hilscher, J., & Szilagyi, J. (2008). In search of distress risk. Journal of Finance, 63(6), 2899–2939
Charitou, A., Neophytou, E., & Charalambous, C. (2004). Predicting corporate failure: Empirical evidence for the UK. European Accounting Review, 13(3), 465–497
Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: Synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 16, 321–357
Chen, S., Härdle, W. K., & Jeong, K. (2010). Forecasting volatility with support vector machine-based GARCH model. Journal of Forecasting, 29(4), 406–433
Choudhary, M. A., & Haider, A. (2012). Neural network models for inflation forecasting: An appraisal. Applied Economics, 44(20), 2631–2635
Cohen, R. B., Polk, C., & Vuolteenaho, T. (2003). The value spread. Journal of Finance, 58(2), 609–641
Dakovic, R., Czado, C., & Berg, D. (2010). Bankruptcy prediction in Norway: A comparison study. Applied Economics Letters, 17(17), 1739–1746
Du Jardin, P. (2018). Failure pattern-based ensembles applied to bankruptcy forecasting. Decision Support Systems, 107, 64–77
Du Jardin, P., Veganzones, D., & Séverin, E. (2019). Forecasting corporate bankruptcy using accrual-based models. Computational Economics, 54(1), 7–43
Duan, J., Sun, J., & Wang, T. (2012). Multiperiod corporate default prediction—A forward intensity approach. Journal of Econometrics, 170(1), 191–209
Falavigna, G. (2012). Financial ratings with scarce information: A neural network approach. Expert Systems with Applications, 39(2), 1784–1792
Figlewski, S., Frydman, H., & Liang, W. (2012). Modeling the effect of macroeconomic factors on corporate default and credit rating transitions. International Review of Economics and Finance, 21(1), 87–105
Foreman, R. D. (2003). A logistic analysis of bankruptcy within the US local telecommunications industry. Journal of Economics and Business, 55(2), 135–166
García, V., Marqués, A. I., Sánchez, J. S., & Ochoa-Domínguez, H. J. (2019). Dissimilarity-based linear models for corporate bankruptcy prediction. Computational Economics, 53(3), 1019–1031
Glover, B. (2016). The expected cost of default. Journal of Financial Economics, 119(2), 284–299
Härdle, W., Lee, Y. J., Schäfer, D., & Yeh, Y. R. (2009). Variable selection and oversampling in the use of smooth support vector machines for predicting the default risk of companies. Journal of Forecasting, 28(6), 512–534
Herbrich, R., Keilbach, M., Graepel, T., Bollmann-Sdorra, P., & Obermayer, K., (1999). Neural networks in economics. In Computational Techniques for Modelling Learning in Economics (pp. 169–196). Springer, Boston, MA
Hinton, G. E., Osindero, S., & Teh, Y.-W. (2006). A fast learning algorithm for deep belief nets. Neural Computation, 18(7), 1527–1554
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780
Huang, W., Nakamori, Y., & Wang, S. Y. (2005). Forecasting stock market movement direction with support vector machine. Computers & Operations Research, 32(10), 2513–2522
James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An introduction to statistical learning. (Vol. 112)Springer.
Jessen, C., & Lando, D. (2015). Robustness of distance-to-default. Journal of Banking & Finance, 50, 493–505
Kim, H., Cho, H., & Ryu, D. (2018). An empirical study on credit card loan delinquency. Economic Systems, 42(3), 437–449
Kim, H., Cho, H., & Ryu, D. (2019). Default risk characteristics of construction surety bonds. Journal of Fixed Income, 29(1), 77–87
Kim, H., Cho, H., & Ryu, D. (2020). Corporate default predictions using machine learning: Literature review. Sustainability, 12(16), 6325
Kim, H., Cho, H., & Ryu, D. (2021). Forecasting consumer credit recovery failure: Classification approaches. Journal of Credit Risk, Forthcoming.
Kuan, C.-M., & Liu, T. (1995). Forecasting exchange rates using feedforward and recurrent neural networks. Journal of Applied Econometrics, 10(4), 347–364
Kukuk, M., & Rönnberg, M. (2013). Corporate credit default models: A mixed logit approach. Review of Quantitative Finance and Accounting, 40(3), 467–483
Lee, Y.-C. (2007). Application of support vector machines to corporate credit rating prediction. Expert Systems with Applications, 33(1), 67–74
Nam, C., Kim, T., Park, N., & Lee, H. (2008). Bankruptcy prediction using a discrete-time duration model incorporating temporal and macroeconomic dependencies. Journal of Forecasting, 27(6), 493–506
Nelson, D. M., Pereira, A. C., & de Oliveira, R. A. (2017). Stock market's price movement prediction with LSTM neural networks. In 2017 International Joint Conference on Neural Networks, IEEE, 1419–1426
Odom, M. D., & Sharda, R. (1990). A neural network model for bankruptcy prediction. In 1990 IJCNN International Joint Conference on Neural Networks, IEEE, 163–168.
Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research, 18(1), 109–131
Pan, Y., Wang, T. Y., & Weisbach, M. S. (2018). How management risk affects corporate debt. Review of Financial Studies, 31(9), 3491–3531
Piri, S., Delen, D., & Liu, T. (2018). A synthetic informative minority over-sampling (SIMO) algorithm leveraging support vector machine to enhance learning from imbalanced datasets. Decision Support Systems, 106, 15–29
Rather, A. M., Agarwal, A., & Sastry, V. N. (2015). Recurrent neural network and a hybrid model for prediction of stock returns. Expert Systems with Applications, 42(6), 3234–3241
Selvin, S., Vinayakumar, R., Gopalakrishnan, E. A., Menon, V. K., & Soman, K. P. (2017). Stock price prediction using LSTM, RNN and CNN-sliding window model. In 2017 International Conference on Advances in Computing, Communications and Informatics, IEEE, 1643–1647
Shumway, T. (2001). Forecasting bankruptcy more accurately: A simple hazard model. Journal of Business, 74(1), 101–124
Siami-Namini, S., & Namin, A. S. (2018). Forecasting economics and financial time series: ARIMA vs. LSTM. arXiv:1803.06386
Tian, S., Yu, Y., & Guo, H. (2015). Variable selection and corporate bankruptcy forecasts. Journal of Banking & Finance, 52, 89–100
Traczynski, J. (2017). Firm default prediction: A Bayesian model-averaging approach. Journal of Financial and Quantitative Analysis, 52(3), 1211–1245
Trustorff, J.-H., Konrad, P. M., & Leker, J. (2011). Credit risk prediction using support vector machines. Review of Quantitative Finance and Accounting, 36, 565–581
Veganzones, D., & Séverina, E. (2018). An investigation of bankruptcy prediction in imbalanced datasets. Decision Support Systems, 112, 111–124
Wilson, R. L., & Sharda, R. (1994). Bankruptcy prediction using neural networks. Decision Support Systems, 11(5), 545–557
Yang, Z., Platt, M. B., & Platt, H. D. (1999). Probabilistic neural networks in bankruptcy prediction. Journal of Business Research, 44(2), 67–74
Zhou, L. (2013). Performance of corporate bankruptcy prediction models on imbalanced dataset: The effect of sampling methods. Knowledge-Based Systems, 41, 16–25
Zmijewski, M. E. (1984). Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research, 22, 59–82
Funding
This work was supported by the 2020 Yeungnam University Research Grant.
Author information
Authors and Affiliations
Contributions
Proposal & original idea, H.K. and D.R.; conceptualization, H.K. and H.C.; modeling, H.K. and D.R.; methodology, H.K. and H.C.; validation, D.R.; resources, H.C.; software, H.K.; literature review, H.K., H.C., and D.R.; economic & business implication, D.R.; writing—original draft preparation, H.K., H.C., and D.R.; writing—review & editing, H.K. and D.R.; discussion, H.K. and D.R.; project administration, D.R. All authors have read and agreed to the published version of the manuscript.
Corresponding author
Ethics declarations
Conflict of interest
There is no conflict of interest regarding the publication of this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In this study, we use five representative classification performance measures to assess each models’ prediction performance. The first four measures—accuracy, precision, recall, and the F1 score—are calculated as follows:
where TP stands for true positive and represents the number of non-bankrupt firms classified as non-bankrupt, TN stands for true negative and represents the number of bankrupt firms classified as bankrupt, FP stands for false positive and represents the number of bankrupt firms classified as non-bankrupt, and FN stands for false negative and represents the number of non-bankrupt firms classified as bankrupt. The fifth measure, the area under the receiver operating characteristic curve, is the area under the graph with the recall on the Y-axis and one minus the specificity on the X-axis, where the specificity is calculated as TN/(FP + TN).
Rights and permissions
About this article
Cite this article
Kim, H., Cho, H. & Ryu, D. Corporate Bankruptcy Prediction Using Machine Learning Methodologies with a Focus on Sequential Data. Comput Econ 59, 1231–1249 (2022). https://doi.org/10.1007/s10614-021-10126-5
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10614-021-10126-5