Modeling and Forecasting Natural Gas Futures Prices Dynamics: An Integrated Approach

We explore and test the capabilities of B-Splines and Dynamic De Rezende-Ferreira five–factor model to replicate the main dynamics and stylized facts of futures curves in the Natural Gas Futures market. Furthermore, we discuss the joint use of these models with a Nonlinear Autoregressive Neural Network for parameters fine–tuning to forecast futures curves. The simulation study highlighted the effectiveness of the proposed framework; empirical results show that the joint use of B–Splines and neural networks provides highest overall performances on the Natural Gas futures market.