Abstract
Understanding the survival prospects of a given population is essential in multiple research and policy areas, including public and private health care and social care, demographic analysis, pension systems evaluation, the valuation of life insurance and retirement income contracts, and the pricing and risk management of novel longevity-linked capital market instruments. This paper conducts a backtesting analysis to assess the predictive performance of Recurrent Neural Networks (RNN) with Gated Recurrent Unit (GRU) architecture in modelling and multivariate time series forecasting of age-specific mortality rates on Chilean mortality data. We investigate the best specification for one, two, and three hidden layers GRU networks and compare the RNN’s forecasting accuracy with that produced by principal component methods, namely a Regularized Singular Value Decomposition (RSVD) model. The empirical results suggest that the forecasting accuracy of RNN models critically depends on hyperparameter calibration and that the two hidden layer RNN-GRU networks outperform the RSVD model. RNNs can generate mortality schedules that are biologically plausible and fit well the mortality schedules across age and time. However, further investigation is necessary to confirm the superiority of deep learning methods in forecasting human survival across different populations and periods.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
- 2.
Due to space constraints, the results for the male population are not included in the main manuscript but are available from the authors upon request.
References
Alho J, Bravo JM, Palmer E (2013). Annuities and life expectancy in NDC. In: Holzmann R, Palmer E, Robalino D (eds) Nonfinancial defined contribution pension schemes in a changing pension world, vol 2. Gender, Politics, and Financial Stability. World Bank Publications, pp 395–436. https://doi.org/10.1596/9780821394786_CH22
Ashofteh A, Bravo JM (2020) A study on the quality of Novel Coronavirus (Covid-19) official datasets. Stat J IAOS 36(2):291–301. https://doi.org/10.3233/SJI-200674
Ashofteh A, Bravo JM (2021) Life table forecasting in COVID-19 times: an ensemble learning approach. In: CISTI 2021 - 16th Iberian conference on information systems and technologies, pp 1–6. https://doi.org/10.23919/CISTI52073.2021.9476583
Ayuso M, Bravo JM, Holzmann R (2021) Getting life expectancy estimates right for pension policy: period versus cohort approach. J Pension Econ Finance 20(2):212–231. https://doi.org/10.1017/S1474747220000050
Ayuso M, Bravo JM, Holzmann R, Palmer E (2021) Automatic indexation of pension age to life expectancy: when policy design matters. Risks 9(5):96. https://doi.org/10.3390/risks9050096
Blake D, Cairns AJG, Dowd K, Kessler AR (2019) Still living with mortality: the longevity risk transfer market after one decade. Br Actuar J 24:1–80
Bravo JM (2016) Taxation of pensions in Portugal: a semi-dual income tax system. CESifo DICE Rep J Inst Comp 14(1):14–23
Bravo JM (2019) Funding for longer lives: retirement wallet and risk-sharing annuities. Ekonomiaz 96(2):268–291
Bravo JM (2021) Pricing participating longevity-linked life annuities: a Bayesian Model Ensemble approach. Eur Actuar J. https://doi.org/10.1007/s13385-021-00279-w
Bravo JM (2021b) Pricing survivor bonds with affine-jump diffusion stochastic mortality models. In: 2021b the 5th international conference on E-commerce, E-business and E-government (ICEEG 2021). Association for Computing Machinery (ACM), New York, pp 91–96. https://doi.org/10.1145/3466029.3466037
Bravo JM (2021c) Forecasting longevity for financial applications: a first experiment with deep learning methods. In: Kamp M et al (eds) Machine Learning and Principles and Practice of Knowledge Discovery in Databases. ECML PKDD 2021. Communications in Computer and Information Science, vol 1525. Springer, Cham. pp. 232–249. https://doi.org/10.1007/978-3-030-93733-1_17
Bravo JM (2021d) Forecasting mortality rates with Recurrent Neural Networks: a preliminary investigation using Portuguese data. In: CAPSI 2021d proceedings (Atas da 21ª Conferência da Associação Portuguesa de Sistemas de Informação 2021), in press
Bravo JM, Ayuso M (2020) Mortality and life expectancy forecasts using Bayesian model combinations: an application to the Portuguese population. RISTI - Revista Ibérica de Sistemas e Tecnologias de Informação E40:128–144. https://doi.org/10.17013/risti.40.128-145
Bravo JM, Ayuso M (2021a) Forecasting the retirement age: a Bayesian model ensemble approach. In: Advances in intelligent systems and computing, AIST, vol 1365. Springer, Cham, pp 123–135 [2021a World Conference on Information Systems and Technologies, WorldCIST 2021]. https://doi.org/10.1007/978-3-030-72657-7_12
Bravo JM, Ayuso M (2021b) Linking pensions to life expectancy: tackling conceptual uncertainty through bayesian model averaging. Mathematics 9(24): 3307, p. 1-27. https://doi.org/10.3390/math9243307
Bravo JM, El Mekkaoui de Freitas N (2018) Valuation of longevity-linked life annuities. Insur Math Econ 78:212–229
Bravo JM, Herce JA (2022) Career breaks, broken pensions? Long-run effects of early and late-career unemployment spells on pension entitlements. J Pension Econ Financ 21(2): 191–217.https://doi.org/10.1017/S1474747220000189
Bravo JM, Nunes JPV (2021) Pricing longevity derivatives via Fourier transforms. Insur Math Econ 96:81–97
Bravo JM, Silva CM (2006) Immunization using a stochastic process independent multifactor model: the Portuguese experience. J Bank Finance 30(1):133–156
Bravo JM, Ayuso M, Holzmann R, Palmer E (2021) Addressing the life expectancy gap in pension policy. Insur Math Econ 99:200–221. https://doi.org/10.1016/j.insmatheco.2021.03.025
Chamboko R, Bravo JM (2016) On the modelling of prognosis from delinquency to normal performance on retail consumer loans. Risk Manage 18(4):264–287
Chamboko R, Bravo JM (2020) A multi-state approach to modelling intermediate events and multiple mortgage loan outcomes. Risks 8:64
Cho K, Van Merriënboer B, Bahdanau D, Bengio Y (2014) On the properties of neural machine translation: encoder-decoder approaches. arXiv preprint arXiv:1409.1259
Coughlan GD, Epstein D, Honig P (2007) Q-forwards: derivatives for transferring longevity and mortality risks. Working paper. J. P. Morgan Pension Advisory Group, London
Deprez P, Shevchenko P, Wüthrich M (2017) Machine learning techniques for mortality modeling. Eur Actuar J 7:337–352
Dowd K, Cairns A, Blake D, Coughlan G, Epstein D, Khalaf-Allah M (2010) Backtesting stochastic mortality models. North Am Act J 14(3):281–298
Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9(8):1735–1780
Hong WH, Yap JH, Selvachandran G et al (2021) Forecasting mortality rates using hybrid Lee-Carter model, artificial neural network and random forest. Complex Intell Syst 7:163–189
Huang JZ, Shen H, Buja A (2009) The analysis of two-way functional data using two-way regularized singular value decompositions. J Am Stat Assoc 104(488):1609–1620
Human Mortality Database (2021) University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany)
Hyndman RJ, Booth H, Yasmeen F (2013) Coherent mortality forecasting: the product-ratio method with functional time series models. Demography 50(1):261–283
Kontis V, Bennett J, Mathers C, Li G, Foreman K, Ezzati M (2017) Future life expectancy in 35 industrialised countries: projections with a Bayesian model ensemble. Lancet 389(10076):1323–1335
Lee RD, Carter L (1992) Modeling and forecasting U.S. mortality. J Am Stat Assoc 87:659–671
Richman R, Wüthrich M (2019) Lee and Carter go machine learning: recurrent neural networks. SSRN https://ssrn.com/abstract=3441030. Accessed 10 Jan 2021
Simões C, Oliveira L, Bravo JM (2021) Immunization strategies for funding multiple inflation-linked retirement income benefits. Risks 9(4):60. https://doi.org/10.3390/risks9040060
Zhang A, Lipton Z, Li M, Smola A (2021) Dive into deep learning. arXiv:2106.11342.
Zhang L, Shen H, Huang JZ (2013) Robust regularized singular value decomposition with application to mortality data. Ann Appl Stat 7(3):1540–1561
Acknowledgements
The authors express their gratitude to the editors and the anonymous referees for his or her careful review and insightful comments, which helped strengthen the quality of the paper. The authors were supported by Portuguese national funds through FCT under the project UIDB/04152/2020—Centro de Investigação em Gestão de Informação (MagIC) and grant UIDB/00315/2020 (BRU-ISCTE).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Bravo, J.M., Santos, V. (2022). Backtesting Recurrent Neural Networks with Gated Recurrent Unit: Probing with Chilean Mortality Data. In: Garcia, M.V., Fernández-Peña, F., Gordón-Gallegos, C. (eds) Advances and Applications in Computer Science, Electronics, and Industrial Engineering. CSEI 2021. Lecture Notes in Networks and Systems, vol 433. Springer, Cham. https://doi.org/10.1007/978-3-030-97719-1_9
Download citation
DOI: https://doi.org/10.1007/978-3-030-97719-1_9
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-97718-4
Online ISBN: 978-3-030-97719-1
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)