Evaluating Forecast Distributions in Neural Network Lee-Carter Type Model for Mortality Rate

In this paper we propose the use of a single hidden layer feed forward artificial neural network as a tool to appropriately capture the nonlinear dynamics of the mortality rates modeled by a Lee-Carter type model. The proposed procedure makes it possible to obtain point forecasts and, by using a bootstrap scheme, the forecast distributions, which allow to take into account the uncertainty of models’ predictions. Empirical evidence on Italian data shows a significant improvement contribution of the proposed methodology.