Analyzing Shapley Additive Explanations to Understand Anomaly Detection Algorithm Behaviors and Their Complementarity

2026-01-30

AI Summary

This work addresses the redundancy commonly observed in unsupervised anomaly detection ensembles, where constituent detectors often rely on similar decision rationales, limiting their complementarity. To overcome this, the study introduces, for the first time, the use of SHAP explanations to quantify each detector’s feature importance assignments, thereby constructing an “explanation profile” to measure inter-model similarity. Based on this, a novel model selection criterion grounded in explanation diversity is proposed. Both theoretical analysis and empirical evaluation demonstrate a strong correlation between explanation dissimilarity and detection complementarity. Ensembles formed by combining high-performing individual detectors with diverse explanation profiles consistently outperform state-of-the-art methods across multiple benchmark datasets, confirming the effectiveness and novelty of the proposed strategy.

Technology Category

• Machine Learning: Ensemble Methods
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Unsupervised anomaly detection is a challenging problem due to the diversity of data distributions and the lack of labels. Ensemble methods are often adopted to mitigate these challenges by combining multiple detectors, which can reduce individual biases and increase robustness. Yet building an ensemble that is genuinely complementary remains challenging, since many detectors rely on similar decision cues and end up producing redundant anomaly scores. As a result, the potential of ensemble learning is often limited by the difficulty of identifying models that truly capture different types of irregularities. To address this, we propose a methodology for characterizing anomaly detectors through their decision mechanisms. Using SHapley Additive exPlanations, we quantify how each model attributes importance to input features, and we use these attribution profiles to measure similarity between detectors. We show that detectors with similar explanations tend to produce correlated anomaly scores and identify largely overlapping anomalies. Conversely, explanation divergence reliably indicates complementary detection behavior. Our results demonstrate that explanation-driven metrics offer a different criterion than raw outputs for selecting models in an ensemble. However, we also demonstrate that diversity alone is insufficient; high individual model performance remains a prerequisite for effective ensembles. By explicitly targeting explanation diversity while maintaining model quality, we are able to construct ensembles that are more diverse, more complementary, and ultimately more effective for unsupervised anomaly detection.

Problem

• unsupervised anomaly detection
• ensemble methods
• model complementarity
• explanation diversity
• anomaly score redundancy

Innovation

• SHAP
• anomaly detection
• ensemble learning
• explanation diversity
• feature attribution