Factor-Based Framework for Multivariate and Multi-step-ahead Forecasting of Large Scale Time Series

State-of-the-art multivariate forecasting methods are restricted to low dimensional tasks, linear dependencies and short horizons. The technological advances (notably the Big data revolution) instead shifting focus problems characterized by a large number of variables, non-linear long In last few years, majority best performing techniques for have been based on deep-learning models. However, such models high requirements in terms availability computational resources suffer from lack interpretability. To cope with limitations these methods, we propose an extension DFML framework, hybrid technique inspired Dynamic Factor Model (DFM) approach, successful methodology econometrics. This improves capabilities DFM implementing assessing both factor estimation as well model-driven data-driven techniques. We assess several method integrations within DFML, show that proposed provides competitive results accuracy efficiency multiple very large-scale (>10 2 variables > 10 3 samples) real tasks.