Thesis title: New perspectives in longevity: methods and insights to understand untracked mortality dynamics
Over the last decades, developed countries have experienced an important mortality decline at all ages, which has been determined by different factors, such as medical progress, healthier lifestyles, and better living conditions. Rising in longevity represents an enormous achievement but it poses major challenges to health care and social security systems.
Some aspects of modern society are planned according to human mortality reliable forecasts, thus understanding longevity dynamics and anticipating mortality changes is crucial for researchers and policymakers.
Despite the great mathematical advances in mortality modeling, canonical methodologies have constantly underestimated future gains in life expectancy during recent decades.
Thereby, this work aims at providing new insights to the analysis and forecasting of human mortality, detecting latent and untracked longevity behaviors. Using recent innovations from Deep Learning, the present dissertation offers different perspectives on mortality developments, introducing mathematical innovations that demographers have not yet explored.
The contribution provides an efficient way to overcome the main limitation of the canonical extrapolative models; providing reliable predictions for unexpected changes in mortality.