Dynamics is what you need for time-series forecasting!
This work addresses the problem of improving time-series forecasting accuracy for researchers and practitioners, but it appears incremental as it builds on existing models by adding a dynamics block.
The paper tackles the challenge of time-series forecasting by hypothesizing that models need to learn underlying data dynamics, and it validates this through a new nomenclature and experiments, showing that incorporating a learnable dynamics block as the final predictor improves performance.
While boundaries between data modalities are vanishing, the usual successful deep models are still challenged by simple ones in the time-series forecasting task. Our hypothesis is that this task needs models that are able to learn the data underlying dynamics. We propose to validate it through both systemic and empirical studies. We develop an original $\texttt{PRO-DYN}$ nomenclature to analyze existing models through the lens of dynamics. Two observations thus emerged: $\textbf{1}$. under-performing architectures learn dynamics at most partially, $\textbf{2}$. the location of the dynamics block at the model end is of prime importance. We conduct extensive experiments to confirm our observations on a set of performance-varying models with diverse backbones. Results support the need to incorporate a learnable dynamics block and its use as the final predictor.