Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MIXAD: Memory-Induced Explainable Time Series Anomaly Detection

Minha Kim, Kishor Kumar Bhaumik, Amin Ahsan Ali, Simon S. Woo

TL;DR

This work introduces MIXAD (Memory-Induced Explainable Time Series Anomaly Detection), a model designed for interpretable anomaly detection that leverages a memory network alongside spatiotemporal processing units to understand the intricate dynamics and topological structures inherent in sensor relationships.

Abstract

For modern industrial applications, accurately detecting and diagnosing anomalies in multivariate time series data is essential. Despite such need, most state-of-the-art methods often prioritize detection performance over model interpretability. Addressing this gap, we introduce MIXAD (Memory-Induced Explainable Time Series Anomaly Detection), a model designed for interpretable anomaly detection. MIXAD leverages a memory network alongside spatiotemporal processing units to understand the intricate dynamics and topological structures inherent in sensor relationships. We also introduce a novel anomaly scoring method that detects significant shifts in memory activation patterns during anomalies. Our approach not only ensures decent detection performance but also outperforms state-of-the-art baselines by 34.30% and 34.51% in interpretability metrics.

MIXAD: Memory-Induced Explainable Time Series Anomaly Detection

TL;DR

This work introduces MIXAD (Memory-Induced Explainable Time Series Anomaly Detection), a model designed for interpretable anomaly detection that leverages a memory network alongside spatiotemporal processing units to understand the intricate dynamics and topological structures inherent in sensor relationships.

Abstract

For modern industrial applications, accurately detecting and diagnosing anomalies in multivariate time series data is essential. Despite such need, most state-of-the-art methods often prioritize detection performance over model interpretability. Addressing this gap, we introduce MIXAD (Memory-Induced Explainable Time Series Anomaly Detection), a model designed for interpretable anomaly detection. MIXAD leverages a memory network alongside spatiotemporal processing units to understand the intricate dynamics and topological structures inherent in sensor relationships. We also introduce a novel anomaly scoring method that detects significant shifts in memory activation patterns during anomalies. Our approach not only ensures decent detection performance but also outperforms state-of-the-art baselines by 34.30% and 34.51% in interpretability metrics.

Paper Structure

This paper contains 19 sections, 8 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Method
  4. Problem Formulation
  5. Spatiotemporal Recurrent Convolution Unit
  6. Memory-Augmented Graph Structure Learning
  7. Memory-Induced Explainable Anomaly Detection (MIXAD)
  8. Anomaly Scoring
  9. Anomaly Interpretation
  10. Experiments and Analysis
  11. Datasets and Baselines
  12. Evaluation Metrics
  13. Performance Comparisons
  14. Ablation Study
  15. Visualization of Node Embeddings
  16. ...and 4 more sections

Figures (5)

  • Figure 1: Overview of the MIXAD framework: Initially, a sparse graph is constructed by calculating pairwise similarities between memory-based node embeddings. Subsequently, input data is processed through a STRGC-based encoder and decoder for self-reconstruction. Throughout this process, an external memory enhances the encoded feature vector by utilizing an attention mechanism between the original feature vector and the memory.
  • Figure 2: T-SNE visualization of node embeddings from two anomaly segments of the SMD dataset.
  • Figure 3: Heatmap visualization of Pearson correlation coefficients for anomaly scores.
  • Figure 4: Visualization of memory activation and anomaly scores for an anomaly segment in the SMD dataset.
  • Figure 5: Anomaly interpretation performance evaluation on the Exathlon dataset.