Dynamic Modes as Time Representation for Spatiotemporal Forecasting

2025/06/01 by Menglin Kong, Kong, Menglin, Vincent Zhihao Zheng +5 · 1 citation
Computer Science · Earth and Planetary Sciences · Environmental Science · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Meteorological Phenomena and Simulations #Soil Geostatistics and Mapping #Time Series Analysis and Forecasting

paper · pdf · doi:10.48550/arxiv.2506.01212

openalex publication_date 2025/06/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

This paper introduces a data-driven time embedding method for modeling long-range seasonal dependencies in spatiotemporal forecasting tasks. The proposed approach employs Dynamic Mode Decomposition (DMD) to extract temporal modes directly from observed data, eliminating the need for explicit timestamps or hand-crafted time features. These temporal modes serve as time representations that can be seamlessly integrated into deep spatiotemporal forecasting models. Unlike conventional embeddings such as time-of-day indicators or sinusoidal functions, our method captures complex multi-scale periodicity through spectral analysis of spatiotemporal data. Extensive experiments on urban mobility, highway traffic, and climate datasets demonstrate that the DMD-based embedding consistently improves long-horizon forecasting accuracy, reduces residual correlation, and enhances temporal generalization. The method is lightweight, model-agnostic, and compatible with any architecture that incorporates time covariates.

Citations

• Dynamic Mode Decomposition and Its Variants
• Long-Range Transformers for Dynamic Spatiotemporal Forecasting
• Learnable Fourier Features for Multi-Dimensional Spatial Positional Encoding
• FC-GAGA: Fully Connected Gated Graph Architecture for Spatio-Temporal\n Traffic Forecasting
• Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting
• Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains
• Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting
• GMAN: A Graph Multi-Attention Network for Traffic Prediction
• GMAN: A Graph Multi-Attention Network for Traffic Prediction
• Time2Vec: Learning a Vector Representation of Time
• Graph WaveNet for Deep Spatial-Temporal Graph Modeling
• N-BEATS: Neural basis expansion analysis for interpretable time series forecasting
• Time-Delay Observables for Koopman: Theory and Applications
• Attention Is All You Need
• Semi-Supervised Classification with Graph Convolutional Networks
• Probabilistic energy forecasting: Global Energy Forecasting Competition 2014 and beyond
• Multi-Scale Context Aggregation by Dilated Convolutions
• De-biasing the dynamic mode decomposition for applied Koopman spectral analysis of noisy datasets
• Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
• Extracting spatial–temporal coherent patterns in large-scale neural recordings using dynamic mode decomposition
• Sequence to Sequence Learning with Neural Networks
• Satellite remote sensing for applied ecologists: opportunities and challenges
• Long Short-Term Memory

Cited by

• Frequency-Constrained Learning for Long-Term Forecasting

Related

• FAITH: Frequency-domain Attention In Two Horizons for Time Series Forecasting
• How Much Can Time-related Features Enhance Time Series Forecasting?
• TimeSieve: Extracting Temporal Dynamics through Information Bottlenecks
• A Decomposition Modeling Framework for Seasonal Time-Series Forecasting
• On the Integration of Spatial-Temporal Knowledge: A Lightweight Approach to Atmospheric Time Series Forecasting