PRISM: A Hierarchical Multiscale Approach for Time Series Forecasting

ICLR 2026 Conference Submission22520 Authors

Keywords: Time series, Forecasting, Multiscale, Hierarchical, Wavelets

TL;DR: We introduce a lightweight tree-based model that hierarchically decomposes time series across time and frequency to forecast time series.

Abstract: Forecasting is critical in areas such as finance, biology, and healthcare. Despite the progress in the field, making accurate forecasts remains challenging because real-world time series contain both global trends, local fine-grained structure, and features on multiple scales in between. Here, we present a new forecasting method, PRISM (Partitioned Representation for hIerarchical Sequence Modeling), that addresses this challenge through a learnable tree-based partitioning of the signal. At the root of the tree, a global representation captures coarse trends in the signal, while recursive splits reveal increasingly localized views of the signal. At each level of the tree, data are projected onto a time-frequency basis (e.g., wavelets or exponential moving averages) to extract scale-specific features, which are then aggregated across the hierarchy. This design allows the model to jointly capture global structure and local dynamics, enabling both reconstruction and forecasting. Experiments across benchmark datasets show that our method outperforms state-of-the-art methods for forecasting and also requires less runtime and memory. Overall, our results demonstrate that hierarchical time-frequency decomposition provides a lightweight and robust framework for forecasting multivariate time series.

Primary Area: learning on time series and dynamical systems

Submission Number: 22520