文章背景与核心概要

大规模时空预测在城市交通、气象监测以及公共卫生等领域发挥着至关重要的作用，然而现有方法往往面临性能瓶颈，难以取得突破性进展。本文作者敏锐地发现，这些限制通常源于“时空复杂度失配”，并创新性地利用空间和时间熵测量来进行诊断。

为了解决这一问题，研究团队提出了一种可扩展的自适应框架，旨在协调空间和时间特征表征。该方法通过低rank矩阵嵌入压缩空间维度以保留基础结构，同时扩展时间跨度以捕获长距离依赖关系。在城市交通、气象和流行病学数据集上的广泛实验表明，该方法不仅显著提高了预测准确性，还增强了跨域迁移能力，为未来的时空预测任务开辟了新途径。

Dimensional Balance Improves Large Scale Spatiotemporal Prediction Performance

Summary

This paper addresses the performance bottlenecks in large-scale spatiotemporal prediction—a critical task for urban traffic, meteorology, and public health. The authors identify that these limitations often stem from a "spatiotemporal complexity mismatch," which they diagnose using spatial and temporal entropy measures. To resolve this, they introduce a scalable, adaptive framework that harmonizes spatial and temporal feature representations. By utilizing low-rank matrix embedding for spatial compression and extending the temporal horizon to capture long-range dependencies, the proposed method achieves significant accuracy gains and improved cross-domain transferability.

Paper Metadata

• arXiv ID: 2605.18793
• Journal Reference: Neural Networks 205 (2027) 109434
• DOI: 10.1016/j.neunet.2026.109434
• Primary Subject: Machine Learning (cs.LG)
• Secondary Subject: Artificial Intelligence (cs.AI)
• Authors: Jing Chen, Shixiang Pan, Yujie Fan, Haocheng Ye, Haitao Xu, Wenqiang Xu

Abstract

精准的时空模式分析在城市交通、气象和公共卫生监测等领域至关重要。然而，现有方法面临性能瓶颈，通常只能带来微乎其微的性能提升，且往往表现出有限的跨域迁移能力。

我们通过空间和时间熵测量来分析这一瓶颈，将其作为时空复杂度失配的诊断指标，而不是将其视为单纯依赖熵对齐就能带来更好预测的保证。从经验来看，在固定的模型容量预算下，更大的失配往往伴随着更高的预测不确定性。

在这一诊断的指导下，我们提出了一种可扩展的自适应框架，能够协调空间和时间特征表征。通过低秩矩阵嵌入压缩空间维度以保留核心结构，同时扩展时间视界以捕获长距离依赖关系，并缓解由时间异质性引起的累积误差。在城市交通、气象和流行病数据集上进行的广泛实验表明，该方法在所评估的领域中实现了显著的准确性提升和广泛的适用性，表明该框架在当前研究之外的广泛时空任务中具有广阔的前景。

Accurate spatiotemporal pattern analysis is critical in fields such as urban traffic, meteorology, and public health monitoring. However, existing methods face performance bottlenecks, typically yielding only incremental gains and often exhibiting limited cross-domain transferability.

We analyze this bottleneck through spatial and temporal entropy measures, which are used as diagnostic indicators of spatiotemporal complexity mismatch rather than as guarantees that entropy alignment alone yields better forecasting. Empirically, larger mismatch is often accompanied by higher prediction uncertainty, especially under a fixed model-capacity budget.

Guided by this diagnostic, we propose a scalable, adaptive framework that harmonizes spatial and temporal feature representations. Spatial dimensionality is compressed via low-rank matrix embedding to preserve essential structure, while an extended temporal horizon captures long-range dependencies and mitigates cumulative errors arising from temporal heterogeneity. Extensive experiments on urban traffic, meteorological, and epidemic datasets demonstrate substantial accuracy gains and broad applicability across the evaluated domains, suggesting that the framework is promising for a wide range of spatiotemporal tasks beyond the current study.

Resources

• Code Repository: GitHub - ST-Balance
• Full-Text Access: View PDF | HTML (Experimental)
• TeX Source: arXiv Source

Submission History

• v1: 11 May 2026
• v2: 10 Aug 2026
• v3: 13 Aug 2026 (Current)