Beyond Stationarity in Time Series: Discovering Causal Structures and Latent Regimes via Markov Blankets
A paper introduces RCBNB-MB, a causal discovery algorithm for time series that relaxes the assumption of a single, time-consistent causal structure. It identifies latent causal regimes and discovers causal graphs within each regime using Markov blankets. Theoretical guarantees are provided, and experiments on simulated and real-world IT monitoring data validate effectiveness.
The paper presents RCBNB-MB, a novel algorithm for causal discovery in time series that accounts for regime changes. It segments time series into regimes with stable causal structures and uses Markov blankets for robustness. The method is validated on simulated data and real-world IT monitoring data.
RCBNB-MB uses an iterative strategy to segment time series into regimes and discover causal graphs within each regime. It leverages Markov blankets instead of direct parents, providing robustness to errors and preserving predictive information. Theoretical guarantees cover recovery of regime transitions and causal graphs under assumptions.
The algorithm addresses limitations in dynamic systems where causal structures change over time, with potential applications in IT monitoring and other domains with regime shifts. Its robustness to errors may make it suitable for real-world deployment.
The method could improve causal analysis in time series data, enabling better decision-making in industries like IT operations, finance, and healthcare where understanding changing causal relationships is critical.
Next signals include further validation on diverse real-world datasets, potential open-source implementation, and adoption in domains requiring dynamic causal modeling. The paper's publication on arXiv may lead to peer-reviewed publication and community evaluation.