A0434
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: Mixed-causal ARMA processes are known to capture the dynamics of locally explosive behavior, such as bubble assets in finance. However, the limited knowledge of the predictive density of mixed-causal processes, especially during explosive bubble events, complicates their forecast and thus limits their use in practical applications. Given the lack of closed-form formulae for the conditional prediction density (except in special cases), simulation-based and sample-based methods have been proposed in the literature. However, these methods can be computationally expensive and do not accurately capture the dynamics during explosive episodes. Mixture density networks (MDNs) are introduced for forecasting time series that exhibit locally explosive behavior. By incorporating Tukey g-and-h transformations 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. Furthermore, the weighted likelihood emphasizes tail observations and crash events, enabling accurate density estimation in the extreme regions most relevant for risk management. Finally, once trained, the MDN produces near-instantaneous density forecasts. Through extensive Monte Carlo simulations and empirical applications, it is shown that the proposed MDN-based framework delivers superior forecasting performance relative to existing approaches.
