Intervention-Aware Time Series Modeling: Capturing and Evaluating Feature Dependencies
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (25) · 20711-20718
DOI: 10.1609/aaai.v40i25.39209 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Ibrahim Delibasoglu und 3 weiteren Autor:innen.
Abstract (Originalsprache)
Understanding how localized changes in one variable affect others in multivariate time series is essential for diagnostics and decision-making in complex systems. Existing models often fail to capture realistic inter-feature dynamics when simulating "what-if" scenarios, leading to inaccurate or uncorrelated reconstructions. We propose CFORVAE, a variational autoencoder framework that explicitly addresses this limitation by combining temporal decomposition with frequency-domain feature correlation modeling. Our architecture uses a dual-path encoding of trend and seasonal components, each projected into attention-pooled latent spaces, and applies Fourier Neural Operators (FNO) to capture cross-feature dependencies in the spectral domain. This decomposition-correlation design enables component-specific latent manipulation and ensures that local modifications propagate realistically across correlated variables. Through extensive experiments, we show that CFORVAE outperforms state-of-the-art baselines in preserving temporal and feature-level dependencies, especially under adjustment-based reconstructions, making it a powerful tool for interpretable "what-if" analysis and diagnostics.
Zitieren
APA
Delibasoglu, I., Chakraborty, S., Heintz, F., & Tiger, M. (2026). Intervention-Aware Time Series Modeling: Capturing and Evaluating Feature Dependencies. Proceedings of the AAAI Conference on Artificial Intelligence, 40(25), 20711-20718. https://doi.org/10.1609/aaai.v40i25.39209
Vancouver
Delibasoglu I, Chakraborty S, Heintz F, Tiger M. Intervention-Aware Time Series Modeling: Capturing and Evaluating Feature Dependencies. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(25):20711-20718. doi:10.1609/aaai.v40i25.39209
BibTeX
@article{Delibasoglu2026,
author = {Delibasoglu, Ibrahim and Chakraborty, Sanjay and Heintz, Fredrik and Tiger, Mattias},
title = {Intervention-Aware Time Series Modeling: Capturing and Evaluating Feature Dependencies},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {25},
pages = {20711--20718},
doi = {10.1609/aaai.v40i25.39209}
}
EUVIMED