Dimension-Reduction with Forward Neural Nets Applied to Mortality

  • Michel Denuit
  • Donatien Hainaut
  • Julien Trufin
Part of the Springer Actuarial book series (SPACT)


In this chapter, we study a particular type of neural networks that are designed for providing a representation of the input with a reduced dimensionality. These networks contains a hidden layer, called bottleneck, that contains a few nodes compared to the previous layers. The output signals of neurons in the bottleneck carry a summarized information that aggregates input signals in a non-linear way. Bottleneck networks offer an interesting alternative to principal component analysis (PCA) or non-linear PCA. In actuarial sciences, these networks can be used for understanding the evolution of longevity during the last century. We also introduce in this chapter a genetic algorithm for calibrating the neural networks. This method combined with a gradient descent speeds up the calibration.


  1. Abdulkarim SA, Garko AB (2015) Forecasting maternal mortality rate using particle Swarm optimization based artificial neural network. Dutse J Pure Appl Sci 1(1):55–59Google Scholar
  2. Antonio K, Bardoutsos A, Ouburg W (2015) Bayesian Poisson log-bilinear models for mortality projections with multiple populations. Eur Actuar J 5:245–281MathSciNetCrossRefGoogle Scholar
  3. Atsalakis G, Nezis D, Matalliotakis G, Ucenic CI, Skiadas C (2007) Forecasting mortality rate using a neural network with fuzzy inference system. Working paper 0806. University of CreteGoogle Scholar
  4. Brouhns N, Denuit M, Vermunt JK (2002) A Poisson log-bilinear regression approach to the construction of projected lifetables. Insurance Math Econom 31(3):373–393MathSciNetCrossRefGoogle Scholar
  5. Cairns AJC (2008) Modelling and management of mortality risk: a review. Scand Actuar J 2–3:79–113MathSciNetCrossRefGoogle Scholar
  6. Currie ID (2016) On fitting generalized linear and non-linear models of mortality. Scand Actuar J 4:356–383MathSciNetCrossRefGoogle Scholar
  7. Cybenko G (1989) Approximation by superpositions of a sigmoidal function. Math Control Signals Syst 2:303–314MathSciNetCrossRefGoogle Scholar
  8. Denuit M, Hainaut D, Trufin J (2019) Effective statistical learning methods for actuaries: GLMs and extensions. Springer, BerlinCrossRefGoogle Scholar
  9. Dong D, McAvoy TJ (1996) Nonlinear principal component analysis—based on principal curves and neural networks. Comp Chem Eng 20:65–78CrossRefGoogle Scholar
  10. Fotheringhame D, Baddeley R (1997) Nonlinear principal components analysis of neuronal spike train data. Biol Cybern 77:282–288CrossRefGoogle Scholar
  11. Fung MC, Peters GW, Shevchenko P (2017) A unified approach to mortality modelling using state-space framework: characterisation, identification, estimation and forecasting. Ann Actuar Sci 2(11):343–389CrossRefGoogle Scholar
  12. Hainaut D (2012) Multidimensional Lee–Carter model with switching mortality processes. Insurance Math Econom 5(2):236–246MathSciNetCrossRefGoogle Scholar
  13. Hainaut D (2018) A neural network analyzer for mortality forecast. ASTIN Bull 48(2):481–450MathSciNetCrossRefGoogle Scholar
  14. Hastie T, Stuetzle W (1989) Principal curves. J Am Stat Assoc 84(406):502–516MathSciNetCrossRefGoogle Scholar
  15. Hornik K (1991) Approximation capabilities of multilayer feedforward networks. Neural Netw 4(2):251–257MathSciNetCrossRefGoogle Scholar
  16. Kramer MA (1991) Nonlinear principal component analysis using autoassociative neural networks. AIChE J 37:233–243CrossRefGoogle Scholar
  17. Lee RD (2000) The Lee-Carter Method for forecasting mortality, with various extensions and applications. North Am Actuar J 4(1):80–91MathSciNetCrossRefGoogle Scholar
  18. Lee RD, Carter L (1992) Modelling and forecasting the time series of US mortality. J Am Stat Assoc 87:659–671Google Scholar
  19. Malthouse EC (1998) Limitations of nonlinear PCA as performed with generic neural networks. IEEE Trans Neural Netw 9:165–173CrossRefGoogle Scholar
  20. McNelis PD (2005) Neural networks in finance: gaining predictive edge in the market. Elsevier, AmsterdamGoogle Scholar
  21. Monahan HA (2000) Nonlinear principal component analysis by neural networks: theory and application to the Lorenz system. J Climate 13:821–835CrossRefGoogle Scholar
  22. Pitacco E (2004) Survival models in a dynamic context: a survey. Insurance Math Econom 35:279–298MathSciNetCrossRefGoogle Scholar
  23. Pitacco E, Denuit M, Haberman S, Olivieri A (2009) Modeling longevity dynamics for pensions and annuity business. Oxford University Press, LondonzbMATHGoogle Scholar
  24. Puddu PE, Menotti A (2009) Artificial neural network versus multiple logistic function to predict 25-year coronary heart disease mortality in the seven countries. Eur J Cardiovasc 16(5):583–591CrossRefGoogle Scholar
  25. Puddu PE, Menotti A (2012) Artificial neural networks versus proportional hazards Cox models to predict 45-year all-cause mortality in the Italian Rural Areas of the Seven Countries Study. BMC Med Res Methodol 12:100CrossRefGoogle Scholar
  26. Renshaw AE, Haberman S (2003) Lee-Carter mortality forecasting with age-specific enhancement. Insurance Math Econom 33:255–272MathSciNetCrossRefGoogle Scholar
  27. Renshaw A, Haberman S (2006) A cohort-based extension to the Lee–Carter model for mortality reduction factors. Insurance Math Econom 38:556–570CrossRefGoogle Scholar
  28. Wilmoth JR (1993) Computational methods for fitting and extrapolating the Lee-Carter model of mortality change. Technical Report; Department of Demography, University of California, BerkeleyGoogle Scholar
  29. Wong-Fupuy C, Haberman (2004) Projecting mortality trends: recent developments in the UK and the US. North Am Actuar J 8:56–83Google Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Michel Denuit
    • 1
  • Donatien Hainaut
    • 2
  • Julien Trufin
    • 3
  1. 1.Université Catholique LouvainLouvain-la-NeuveBelgium
  2. 2.Université Catholique de LouvainLouvain-la-NeuveFrance
  3. 3.Université Libre de BruxellesBrusselsBelgium

Personalised recommendations