Kernel Outlier Detection

2025-06-28

AI Summary

Existing high-dimensional anomaly detection methods often rely on strong distributional assumptions and suffer from sensitivity to hyperparameter tuning. To address these limitations, this paper proposes Kernelized Projection Pursuit (KPP), a distribution-free and hyperparameter-robust approach. KPP employs kernel-induced nonlinear embeddings to transform input data and introduces a novel ensemble-based directional search strategy that adaptively identifies the most discriminative projection directions in a low-dimensional subspace. Furthermore, it incorporates a multi-direction fusion mechanism to enhance robustness and generalization. Extensive experiments on three structurally complex small-scale datasets and four large-scale benchmark datasets demonstrate that KPP consistently outperforms state-of-the-art methods—achieving superior detection accuracy, exceptional stability across diverse settings, and remarkable insensitivity to hyperparameter choices. By combining interpretability with ease of deployment, KPP establishes a new paradigm for high-dimensional outlier detection.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Kernel Methods
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

A new anomaly detection method called kernel outlier detection (KOD) is proposed. It is designed to address challenges of outlier detection in high-dimensional settings. The aim is to overcome limitations of existing methods, such as dependence on distributional assumptions or on hyperparameters that are hard to tune. KOD starts with a kernel transformation, followed by a projection pursuit approach. Its novelties include a new ensemble of directions to search over, and a new way to combine results of different direction types. This provides a flexible and lightweight approach for outlier detection. Our empirical evaluations illustrate the effectiveness of KOD on three small datasets with challenging structures, and on four large benchmark datasets.

Problem

• Proposes KOD for high-dimensional outlier detection
• Overcomes limitations of existing hyperparameter-dependent methods
• Introduces novel ensemble and combination techniques for flexibility

Innovation

• Kernel transformation for high-dimensional outlier detection
• Ensemble of directions for flexible search
• Lightweight projection pursuit approach