Preference isolation forest for structure-based anomaly detection

Filippo Leveni

doi:10.1016/j.patcog.2025.112405

Preference isolation forest for structure-based anomaly detection

2025-09-12

0

PRE

AI

DOI：10.1016/j.patcog.2025.112405

Original

Original request for help

Save

Abstract

En 中文

• We propose a general framework to detect anomalies with respect to parametric structured patterns. • We detect anomalies by embedding data into a high-dimensional preference space that exploits priors on genuine structures. • We detect anomalies as isolated points in preference space using tailored isolation methods with proper distance measures. • We propose a sliding window extension able to deal with unknown but smooth patterns by leveraging on locality priors. • Extensive experiments on synthetic and real-world data demonstrate the effectiveness of structure-based anomaly detection.

Preference isolation forest for structure-based anomaly detection

Abstract

Journal

Organization

Cited Papers