Dear all,

you're invited to the seminar that will take place, in hybrid mode, at the Department of Statistics and Quantitative Methods, University of Milano-Bicocca. More details:

21st of December 2022, h 16:30, Aula seminari 4026, Building U7, 4th floor Department of Statistics and Quantitative Methods, University of Milano-Bicocca Webex Link: https://unimib.webex.com/unimib/j.php?MTID=mcd5a40a675fd6136e8c3be8a8ead4143

Join by meeting number Meeting number (access code): 2742 148 3807 Meeting password: x3765HJJdRf (93765455 from phones)

Speaker: Salvatore Scognamiglio

Title: ''Deep learning models for longevity risk''

Abstract: In recent decades, the mortality rates of most developed countries have been gradually declining due to improvements in public health, medical advances, lifestyle changes and government regulation. Although longevity improvement is an obvious benefit for society, it could also represent a risk for governments and insurance companies. Indeed, they could get in financial trouble if they do not adequately consider these longevity improvements for retirement planning and life insurance product pricing. The risk that future mortality and life expectancy outcomes turn out differently than expected is typically called longevity risk, and its management requires stochastic mortality projection models. We introduce a neural network (NN) to fit the Lee-Carter models and some of its extensions on multiple populations. The NN architecture is specifically designed to calibrate each individual model using all available information instead of using a population-specific subset of data as in the traditional estimation schemes. An extensive set of numerical experiments performed on all the countries of the Human Mortality Database shows the effectiveness of our approach. In particular, the resulting parameter estimates appear smooth and less sensitive to the random fluctuations often present in the mortality rates’ data, especially for low-population countries. In addition, the forecasting performance results significantly improved as well.

Best regards, Valeria

********************************************** Valeria Bignozzi Professore associato di Metodi Matematici dell’Economia e delle Scienze Attuariali e Finanziarie Dipartimento di Statistica e Metodi Quantitativi (DiSMeQ), Università degli Studi di Milano-Bicocca, Edificio U7