Hybrid short/long-horizon system

Background: Mortality forecasting methods often rely on extrapolation of time-series components (like those from the Lee-Carter model), which can lead to systematic biases if the underlying trend decelerates or changes beyond historically observed rates.

Question / Future Work: A hybrid forecasting system should be developed that strategically combines the momentum-capturing capabilities of time-series models (like ARIMA models used for short-horizon momentum) with the constrained, long-horizon dynamics of the flow-field approach, transitioning from one to the other at an appropriate horizon.