Cira PERNA | Progetti

In this project we explore the use of a single hidden layer feed-forward artificial Neural Networks (NNs) as a forecasting tool to capture the nonlinear dynamics of the mortality rates. The approach is able to obtain valid point forecasts and, by using the bootstrap, the forecast distributions which allow to evaluate how much uncertainty is associated with each point forecast.

In this project we focus on the estimation of forecast distributions in non linear autoregressive time series and, in this context, we propose the use of non parametric techniques. In particular, we evaluate the approach based on feed-forward neural networks (NN) to approximate the original nonlinear process and derive valid point forecasts and the pair bootstrap scheme as a resampling device

In this project we focus on range-based volatility and, in this context, we propose the use of Extreme Learning Machines (ELMs) to appropriately capture the nonlinear dynamics of these volatility measures.