KITE: Knowledge-Guided Probabilistic Modeling for Time Series Forecasting with Exogenous Variables

Hanyin Cheng, Jingrong Zhou, Yang Shu, Chenjuan Guo
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:19211-19235, 2026.

Abstract

Probabilistic forecasting with exogenous variables is vital for decision-making but remains underexplored compared to deterministic methods. We propose KITE, a knowledge-guided probabilistic modeling framework designed to bridge this gap by addressing two key bottlenecks: (1) topological disparity in sampling initialization and (2) spurious covariate correlations during the iterative conditional generation process. KITE introduces a History-Conditional Manifold to construct an informative source distribution from historical dynamics, effectively anchoring the starting point closer to the target space. Additionally, a Knowledge-Guided Conditioning module is developed to regularize variable interactions using statistical priors, suppressing spurious correlations and enhancing the robustness of covariate conditioning. Extensive experiments demonstrate that KITE outperforms state-of-the-art methods in both deterministic and probabilistic forecasting.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-cheng26t, title = {{KITE}: Knowledge-Guided Probabilistic Modeling for Time Series Forecasting with Exogenous Variables}, author = {Cheng, Hanyin and Zhou, Jingrong and Shu, Yang and Guo, Chenjuan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {19211--19235}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/cheng26t/cheng26t.pdf}, url = {https://proceedings.mlr.press/v306/cheng26t.html}, abstract = {Probabilistic forecasting with exogenous variables is vital for decision-making but remains underexplored compared to deterministic methods. We propose KITE, a knowledge-guided probabilistic modeling framework designed to bridge this gap by addressing two key bottlenecks: (1) topological disparity in sampling initialization and (2) spurious covariate correlations during the iterative conditional generation process. KITE introduces a History-Conditional Manifold to construct an informative source distribution from historical dynamics, effectively anchoring the starting point closer to the target space. Additionally, a Knowledge-Guided Conditioning module is developed to regularize variable interactions using statistical priors, suppressing spurious correlations and enhancing the robustness of covariate conditioning. Extensive experiments demonstrate that KITE outperforms state-of-the-art methods in both deterministic and probabilistic forecasting.} }
Endnote
%0 Conference Paper %T KITE: Knowledge-Guided Probabilistic Modeling for Time Series Forecasting with Exogenous Variables %A Hanyin Cheng %A Jingrong Zhou %A Yang Shu %A Chenjuan Guo %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-cheng26t %I PMLR %P 19211--19235 %U https://proceedings.mlr.press/v306/cheng26t.html %V 306 %X Probabilistic forecasting with exogenous variables is vital for decision-making but remains underexplored compared to deterministic methods. We propose KITE, a knowledge-guided probabilistic modeling framework designed to bridge this gap by addressing two key bottlenecks: (1) topological disparity in sampling initialization and (2) spurious covariate correlations during the iterative conditional generation process. KITE introduces a History-Conditional Manifold to construct an informative source distribution from historical dynamics, effectively anchoring the starting point closer to the target space. Additionally, a Knowledge-Guided Conditioning module is developed to regularize variable interactions using statistical priors, suppressing spurious correlations and enhancing the robustness of covariate conditioning. Extensive experiments demonstrate that KITE outperforms state-of-the-art methods in both deterministic and probabilistic forecasting.
APA
Cheng, H., Zhou, J., Shu, Y. & Guo, C.. (2026). KITE: Knowledge-Guided Probabilistic Modeling for Time Series Forecasting with Exogenous Variables. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:19211-19235 Available from https://proceedings.mlr.press/v306/cheng26t.html.

Related Material