Dynamic Modes as Time Representation for Spatiotemporal Forecasting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kong, Menglin, Zheng, Vincent Zhihao, Wang, Xudong, Sun, Lijun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909717856518144
author Kong, Menglin
Zheng, Vincent Zhihao
Wang, Xudong
Sun, Lijun
author_facet Kong, Menglin
Zheng, Vincent Zhihao
Wang, Xudong
Sun, Lijun
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2506_01212
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Dynamic Modes as Time Representation for Spatiotemporal Forecasting
Kong, Menglin
Zheng, Vincent Zhihao
Wang, Xudong
Sun, Lijun
Machine Learning
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.
title Dynamic Modes as Time Representation for Spatiotemporal Forecasting
topic Machine Learning
url https://arxiv.org/abs/2506.01212