PRISM: Hierarchical Time Series Forecasting
Published:Dec 31, 2025 14:51
•1 min read
•ArXiv
Analysis
This paper introduces PRISM, a novel forecasting method designed to handle the complexities of real-world time series data. The core innovation lies in its hierarchical, tree-based partitioning of the signal, allowing it to capture both global trends and local dynamics across multiple scales. The use of time-frequency bases for feature extraction and aggregation across the hierarchy is a key aspect of its design. The paper claims superior performance compared to existing state-of-the-art methods, making it a potentially significant contribution to the field of time series forecasting.
Key Takeaways
- •PRISM is a new time series forecasting method.
- •It uses a hierarchical, tree-based approach to capture both global and local features.
- •It employs time-frequency bases for feature extraction.
- •The method outperforms state-of-the-art methods in experiments.
- •The code is publicly available.
Reference
“PRISM addresses the challenge through a learnable tree-based partitioning of the signal.”