Bayesian importance-weighted support vector data description

doi:10.1016/j.patcog.2026.113567

Bayesian importance-weighted support vector data description

2026-03-22

0

PRE

AI

DOI：10.1016/j.patcog.2026.113567

Original

Original request for help

Save

Abstract

En 中文

• Proposed a Bayesian Importance-Weighted SVDD model that handles multiple anomaly types. • Introduction of importance weighting that prioritizes the detection of severe anomalies. • Probabilistic framework, capturing uncertainty in normal and anomaly data. • The proposed method has higher detection rates and F1 scores on simulated datasets. • The proposed method outperformed existing SVDD methods in detecting fires earlier.

Keywords:

Bayesian importance-weighted SVDD

anomaly detection

severe anomalies

probabilistic framework

F1 score

Bayesian importance-weighted support vector data description

Abstract

Journal

Organization