A1864
Title: Tail-aware density forecasting of locally explosive time series: A neural network approach
Authors: Julien Peignon - Paris Dauphine University (France) [presenting]
Arthur Thomas - Paris Dauphine University - PSL (France)
Elena Dumitrescu - University Paris-Pantheon-Assas (France)
Abstract: A Mixture Density Network specifically designed for forecasting time series that exhibit locally explosive behavior is proposed. By incorporating skewed t-distributions as Mixture components, the approach offers enhanced flexibility in capturing the skewed, heavy-tailed, and potentially multimodal nature of predictive densities associated with bubble dynamics modeled by mixed causal-noncausal ARMA processes. An adaptive weighting scheme is implemented that emphasizes tail observations during training and leads to accurate Density estimation in the extreme regions most relevant for financial applications. Once trained, the MDN produces near-instantaneous Density forecasts. Through extensive Monte Carlo simulations and two empirical applications on natural gas prices and inflation, the proposed MDN-based framework delivers superior forecasting performance relative to existing approaches.
