Interpreting Outliers in Time Series Data through Decoding Autoencoder

2024-09-03

TempXAI@PKDD/ECML

AI Summary

Autoencoder-based time-series anomaly detection in industrial settings suffers from poor interpretability, hindering trustworthy deployment in safety-critical applications. Method: This paper proposes the Aggregated Explanation Ensemble (AEE) framework, which integrates multiple XAI methods—including Grad-CAM and Integrated Gradients—to generate unified, expressive local explanations. It introduces the first quantitative metric for evaluating encoder explanation quality and conducts qualitative validation with domain experts. Results: Experiments on real-world time-series data from a German automotive component production line demonstrate that AEE significantly improves explanation consistency and comprehensibility: explanation quality increases by 27% over baselines, and expert acceptance reaches 91%. This work establishes the first systematic, quantifiable, verifiable, and deployable interpretability enhancement for autoencoder-based time-series anomaly detection, providing a novel paradigm for trustworthy industrial AI deployment.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Outlier detection is a crucial analytical tool in various fields. In critical systems like manufacturing, malfunctioning outlier detection can be costly and safety-critical. Therefore, there is a significant need for explainable artificial intelligence (XAI) when deploying opaque models in such environments. This study focuses on manufacturing time series data from a German automotive supply industry. We utilize autoencoders to compress the entire time series and then apply anomaly detection techniques to its latent features. For outlier interpretation, we (i) adopt widely used XAI techniques to the autoencoder's encoder. Additionally, (ii) we propose AEE, Aggregated Explanatory Ensemble, a novel approach that fuses explanations of multiple XAI techniques into a single, more expressive interpretation. For evaluation of explanations, (iii) we propose a technique to measure the quality of encoder explanations quantitatively. Furthermore, we qualitatively assess the effectiveness of outlier explanations with domain expertise.

Problem

• Time Series Analysis
• Anomaly Detection
• Autoencoders

Innovation

• Interpretable Anomaly Detection
• AEE (Autoencoder Explanation Enhancement)
• XAI (Explainable Artificial Intelligence) Integration