Abstract
Monitoring liquidity management of banks is one of the prime tasks of central banks. A bank that manages its liquidity inadequately can severely harm its liquidity position and potentially threaten the stability of the entire financial system. Central banks try to anticipate these risks by carefully monitoring the liquidity management of banks in large-value payment systems (LVPSs). Typically, they do this based on statistical methods in which various risk indicators related to the liquidity usage of banks are calculated from the transaction log of an LVPS. These indicators need to be manually analyzed by payment experts to find irregularities that could signal potential risks. Although statistical methods provide much insight into the liquidity management of banks, they do not scale well to the large number of banks that are subject to risk monitoring and the high velocity by which payments are nowadays settled. In this paper, we investigate whether the liquidity management of banks can be monitored more efficiently by anomaly detection. We construct different probabilistic classifiers that classify delta sequences of banks by the corresponding bank. A delta sequence captures the change in the liquidity position of a bank in an LVPS throughout a given day. Accordingly, anomalies in the intraday liquidity usage of banks are detected by determining whether the classifiers misclassify recent delta sequences that were not used to train the classifiers. Our results show that recurrent neural networks are well suited to perform this classification task and detect many irregularities in payment behavior that are interesting for the supervisors and operators of an LVPS.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The systemic importance of an FMI is typically determined by the total amount of liquidity that it processes over a given period. As a reference, the Mexican LVPS (SPEI) processed more than 200 times the Mexican gross domestic product in 2015 (Banxico 2016).
Free-riding is a liquidity management strategy in which a bank tends to delay its payments until it received enough liquidity from other banks. Recycling payments in this way provides essentially ‘free’ liquidity since a bank does not need to use its reserves or lend liquidity from external sources.
The delta positions of a bank in a delta sequence are not independent.
There have also been efforts to optimize neural networks using derivative-free optimization, see e.g. (Aly et al. 2019), but this does not resolve the vanishing gradient problem. Vanishing gradients are a property of the RNN’s architecture. Small changes at the beginning of a delta sequence will have a neglectable effect on the network’s output many time intervals ahead, regardless of the chosen optimization method.
See (Alexandrova-Kabadjova et al. 2013) for more details about SPEI.
Originally, the transaction log also included data of several small banks who made only a few payments. We excluded these banks from the analysis because there were not enough payments to learn their normal liquidity management patterns.
There was a bank that had a delta sequence consisting of all zeros (i.e. it had no incoming and outgoing payments at a given day). We removed this sequence.
References
Adachi, K. (2016). Matrix-based introduction to multivariate data analysis. Berlin: Springer.
Alam, P., Booth, D., Lee, K., & Thordarson, T. (2000). The use of fuzzy clustering algorithm and self-organizing neural networks for identifying potentially failing banks: An experimental study. Expert Systems with Applications, 18(3), 185–199.
Alexandrova-Kabadjova, B., & Sols-Robleda, F. (2013). Managing intraday liquidity: The Mexican experience. In Simulation in computational finance and economics: tools and emerging applications, chap. 8 (pp. 143–160). IGI Global.
Allen, F., & Gale, D. (2000). Financial contagion. Journal of Political Economy, 108(1), 1–33.
Aly, A., Guadagni, G., & Dugan, J.B. (2019). Derivative-free optimization of neural networks using local search. In Proceedings of the 10th annual ubiquitous computing, electronics and mobile communication conference (pp. 0293–0299). IEEE.
Banxico. (2016). Interbank electronic payment system (SPEI): Principles for financial market infrastructures disclosure. Technical report, Banco de México. https://www.banxico.org.mx/payment-systems/d/90965A55-8F44-7DD2-45CF-2BF1D7C0B75B.pdf.
BCBS. (2013). Monitoring tools for intraday liquidity management. Technical report, Basel Committee on Banking Supervision.
Bengio, Y., Simard, P., & Frasconi, P. (1994). Learning long-term dependencies with gradient descent is difficult. IEEE Transactions on Neural Networks, 5(2), 157–166.
Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: A survey. ACM Computing Surveys, 41(3), 15.
Chen, W. H., & Shih, J. Y. (2006). A study of Taiwan’s issuer credit rating systems using support vector machines. Expert Systems with Applications, 30(3), 427–435.
Cho, K., Van Merriënboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder–decoder for statistical machine translation. In Proceedings of the 2014 conference on empirical methods in natural language processing (pp. 1724–1734).
Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555.
CPSS. (2012). Principles for financial market infrastructures. Technical report, Bank for International Settlements (BIS) and International Organization of Securities Commissions (IOSCO).
Denbee, E., Garratt, R., & Zimmerman, P. (2012). Identification of over and under provision of liquidity in real-time payment systems. Journal of Financial Market Infrastructures, 4(2), 1–20.
Diehl, M. (2013). Measuring free riding in large-value payment systems: The case of TARGET2. Journal of Financial Market Infrastructures, 1(3), 31–53.
Glorot, X., & Bengio, Y. (2010). Understanding the difficulty of training deep feedforward neural networks. In Proceedings of the 13th international conference on artificial intelligence and statistics (pp. 249–256).
Glowka, M., Paulick, J., & Schultze, I. (2018). The absence of evidence and the evidence of absence: An algorithmic approach for identifying operational outages in TARGET2. Journal of Financial Market Infrastructures, 6(2/3), 63–91.
Heijmans, R., & Heuver, R. (2014). Is this bank ill? The diagnosis of doctor TARGET2. Journal of Financial Market Infrastructures, 2(3), 3–36.
Hinton, G., Srivastava, N., & Swersky, K. (2012). Neural networks for machine learning lecture 6a overview of mini-batch gradient descent. https://www.cs.toronto.edu/~tijmen/csc321/slides/lecture_slides_lec6.pdf.
Hochreiter, S., Bengio, Y., Frasconi, P., & Schmidhuber, J. (2001). Gradient flow in recurrent nets: The difficulty of learning long-term dependencies, chap. 14 (pp. 237–244). Wiley-IEEE Press.
Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780.
Kohavi, R. (1995). A study of cross-validation and bootstrap for accuracy estimation and model selection. In Proceedings of the 14th international joint conference on artificial intelligence (Vol. 2, pp. 1137–1143).
Laeven, L., & Valencia, F. (2013). Systemic banking crises database. IMF Economic Review, 61(2), 225–270.
LeCun, Y., Bottou, L., Orr, G. B., & Müller, K. R. (1998). Efficient backprop, Lecture Notes in Computer Science (Vol. 1524, pp. 9–50). Berlin: Springer.
León, C., Moreno, J. F., & Cely, J. (2016). Whose balance sheet is this? Neural networks for banks’ pattern recognition. Borradores de Economia 959, Banco de la Republica de Colombia.
Maaten, L.v.d., & Hinton, G. (2008). Visualizing data using t-SNE. Journal of Machine Learning Research,9(Nov), 2579–2605.
Olmeda, I., & Fernández, E. (1997). Hybrid classifiers for financial multicriteria decision making: The case of bankruptcy prediction. Computational Economics, 10(4), 317–335.
Pascanu, R., Mikolov, T., & Bengio, Y. (2013). On the difficulty of training recurrent neural networks. In Proceedings of the 30th international conference on machine learning (pp. 1310–1318).
Pasiouras, F., Gaganis, C., & Doumpos, M. (2007). A multicriteria discrimination approach for the credit rating of Asian banks. Annals of Finance, 3(3), 351–367.
Sarkar, S., & Sriram, R. S. (2001). Bayesian models for early warning of bank failures. Management Science, 47(11), 1457–1475.
Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. (2014). Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1), 1929–1958.
Werbos, P. J. (1990). Backpropagation through time: What it does and how to do it. Proceedings of the IEEE, 78(10), 1550–1560.
Acknowledgements
This paper was written while Ron Triepels was visiting Banco de México in Mexico City. We would like to thank Biliana Alexandrova-Kabadjova from Banco de México for her help in developing the case study.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix: Results for Standard z-Normalization
Appendix: Results for Standard z-Normalization
Table 3 shows the performance of the classifiers in the case of normalizing the delta sequences by standard z-normalization. If we compare these results with those for bank-specific normalization reported in Table 2, we see that standard z-normalization harms the performance of the classifiers. This observation can be explained by the heterogeneous nature of banks. Typically, an LVPS consists of a few large banks that have relatively large (positive or negative) delta positions, and many small banks that have relatively small (positive or negative) delta positions. When we use standard z-normalization, the delta positions of large banks dominate the normalization and cause the delta positions of small banks to be rescaled to very small ranges. This makes it much harder to distinguish the delta sequences of small banks.
Rights and permissions
About this article
Cite this article
Triepels, R., Daniels, H. & Berndsen, R. Monitoring Liquidity Management of Banks With Recurrent Neural Networks. Comput Econ 57, 89–112 (2021). https://doi.org/10.1007/s10614-020-10067-5
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10614-020-10067-5