arXiv：cs.AI· Muhammad Hasan Ferdous, Ian Stockwell, Md Osman Gani·· 22 小时前
DCD：面向自相关与非平稳时序数据的基于分解的因果发现框架
DCD: Decomposition-based Causal Discovery from Autocorrelated and Non-Stationary Temporal Data
AI 导读
DCD 是一种基于分解的因果发现框架，将每条时间序列拆分为趋势、季节和残差分量并分别做因果分析：趋势用平稳性检验、季节用核依赖度量、残差用约束型因果发现，再整合为统一的多尺度因果结构。在大量合成基准和真实气候数据上，该框架比现有最优基线更准确地还原真实因果结构，尤其在强非平稳和时间自相关条件下。
正文
Abstract:Multivariate time series in domains such as finance, climate science, and healthcare often exhibit long-term trends, seasonal patterns, and short-term fluctuations, complicating causal inference under non-stationarity and autocorrelation. Existing causal discovery methods typically operate on raw observations, making them vulnerable to spurious edges and misattributed temporal dependencies. We introduce a decomposition-based causal discovery framework that separates each time series into trend, seasonal, and residual components and performs component-specific causal analysis. Trend components are assessed using stationarity tests, seasonal components using kernel-based dependence measures, and residual components using constraint-based causal discovery. The resulting component-level graphs are integrated into a unified multi-scale causal structure. This approach isolates long- and short-range causal effects, reduces spurious associations, and improves interpretability. Across extensive synthetic benchmarks and real-world climate data, our framework more accurately recovers ground-truth causal structure than state-of-the-art baselines, particularly under strong non-stationarity and temporal autocorrelation.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML) |
| MSC classes: | 62M10 |
| Cite as: | arXiv:2602.01433 [cs.LG] |
| (or arXiv:2602.01433v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2602.01433 arXiv-issued DOI via DataCite |
Submission history
From: Muhammad Hasan Ferdous [view email]
[v1]
Sun, 1 Feb 2026 20:45:05 UTC (1,729 KB)
[v2]
Tue, 28 Apr 2026 04:21:44 UTC (4,296 KB)
[v3]
Mon, 12 Oct 2026 23:23:02 UTC (4,279 KB)
来源：arXiv：cs.AI · arxiv.org