Joint Selective State Space Model and Detrending for Robust Time Series Anomaly Detection

TL;DR

This paper introduces a novel anomaly detection method combining a selective state space model for capturing long-term dependencies with a multi-stage detrending mechanism to handle non-stationary data, achieving superior results on real-world datasets.

Contribution

The paper proposes a joint model integrating a selective state space approach with multi-stage detrending for improved robustness in time series anomaly detection.

Findings

Outperforms 12 baseline methods on real-world datasets

Effectively captures long-range dependencies in time series

Mitigates trend effects in non-stationary data

Abstract

Deep learning-based sequence models are extensively employed in Time Series Anomaly Detection (TSAD) tasks due to their effective sequential modeling capabilities. However, the ability of TSAD is limited by two key challenges: (i) the ability to model long-range dependency and (ii) the generalization issue in the presence of non-stationary data. To tackle these challenges, an anomaly detector that leverages the selective state space model known for its proficiency in capturing long-term dependencies across various domains is proposed. Additionally, a multi-stage detrending mechanism is introduced to mitigate the prominent trend component in non-stationary data to address the generalization issue. Extensive experiments conducted on realworld public datasets demonstrate that the proposed methods surpass all 12 compared baseline methods.