Work

An anomaly detection method for conformal mechanism

Year: 2025

Type: article

Abstract: This paper introduces a novel hypersphere-based anomaly detection method enhanced by a conformal mapping transformation to address challenges such as data sparsity, computational complexity, and the scarcity of labeled anomalies in high-dimensional spaces. The core of our approach, termed HS-MT, is the construction of two separate hyperspheres to model normal and anomalous instances, leveraging their efficiency and interpretability. A key contribution is the derivation and theoretical validation of a specific conformal mapping function that projects data into a transformed space, effectively condensing anomalous points into tighter clusters and improving their detectability. Experiments on synthetic and real-world UCI datasets demonstrate that HS-MT consistently outperforms state-of-the-art methods in accuracy, recall, and computational efficiency, particularly in scenarios with high anomaly ratios. The results confirm that the proposed mapping function and dual-hypersphere design together offer a scalable and theoretically grounded solution for identifying subtle anomalies in complex datasets.

Source: Egyptian Informatics Journal

DOI: 10.1016/j.eij.2025.100824

Authors: Jian Zheng, Jingjing Sun, Qiankun Liao, Mengyun Xie, Shengye Wang

Institutions: Chongqing Technology and Business University, Chongqing Polytechnic Institute

Language: English

Topic: Anomaly Detection Techniques and Applications

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $1,200

Funders: Chongqing Technology and Business University, Natural Science Foundation of Chongqing, Natural Science Foundation Project of Chongqing, Chongqing Science and Technology Commission