MSCENet: A Multi-Scale Correlation Enhanced Network for Anomaly Detection

2026-07-07

AI Summary

This work addresses the challenge of simultaneously modeling multi-scale temporal dynamics and complex inter-variable dependencies in multivariate time series anomaly detection by proposing the MSCENet framework. The method innovatively integrates multi-scale dilated convolutions, MixHop graph convolution, and a gating mechanism to adaptively capture spatiotemporal correlations across varying time scales. Furthermore, it jointly learns dynamic dependencies among variables through an adaptive graph structure. Experimental results on real-world datasets—including SMD, PSM, and SWaT—demonstrate that MSCENet significantly outperforms existing state-of-the-art approaches, offering a novel and effective solution for anomaly detection in complex, real-world scenarios.

Abstract

In the field of multivariate time series anomaly detection, against the backdrop of increasing data complexity and complex dependencies across multiple temporal scales, traditional methods often struggle to simultaneously capture temporal dynamic features and intricate inter-series correlations. To address this, we propose an innovative framework, MSCENet, which leverages advanced spatio-temporal learning and multi-scale learning techniques to enhance detection accuracy. MSCENet includes a fine-grained temporal convolution module that captures complex temporal dependencies through dilated convolutions, enabling the detection of both short- and long-term patterns. Additionally, the framework models inter-series relationships as a graph structure, using Mixhop graph convolutions to adaptively capture spatial dependencies across varying time scales. To support robust anomaly detection, the multi-scale gated convolution module in MSCENet integrates spatial and temporal attributes through gated mechanisms, facilitating the detection of subtle variations across multiple scales. Experimental evaluations on real-world datasets: SMD, PSM, and SWaT. It provides an adaptable and high-performance solution for anomaly detection in complex time series data environments.

Problem

• multivariate time series
• anomaly detection
• temporal dependencies
• inter-series correlations
• multi-scale

Innovation

• multi-scale learning
• spatio-temporal graph convolution
• dilated convolution
• gated mechanism
• anomaly detection