Review on Information Fusion‐Based Data Mining for Improving Complex Anomaly Detection

Wiley Interdisciplinary Reviews Data Mining and Knowledge Discovery, Journal Year: 2025, Volume and Issue: 15(2)

Published: May 8, 2025

ABSTRACT Anomaly predicated upon multiple distributed hybrid sensors frequently uses approaches, integrating techniques derived from statistical analysis, probability, data mining, machine learning, deep and signal denoising. Many of these methods are based on the analysis irregularities, continuity, correlation, consistency, aiming to discern anomalous patterns normal behavior. By leveraging information fusion aims enhance situational awareness, detect potential threats or abnormalities, improve decision‐making processes in complex environments. It addresses uncertainties by diverse sources, thereby enhancing performance, reducing dependency individual sensors. This study examines applications single sensor data, revealing common strategies, identifying strengths weaknesses, solutions for detecting diagnosing anomalies analyzing low, large, context homogeneous heterogeneous systems. Information evaluated their performance various levels algorithm complexity. in‐depth bibliographic involved searching top indexing databases such as Web Science Scopus. The inclusion criteria were articles published between 2012 2024. search capitalized specific keywords follows: “sensor malfunction,” anomaly,” failure,” fusion,” “anomaly mining.” Publications that did not strictly focus analytical processing anomaly detection, diagnosis, prognosis excluded. In conclusion, practice promotes transparency elucidating process combining information, enabling multitude perspectives, aligning with established best practices field. Data deviation remains primary criterion using mostly learning extensively techniques. Nevertheless, state‐of‐the‐art algorithms neural networks still require further contextual interpretation analysis. Functional safety intended functionality breaching can lead errors, physical harm, erosion trust autonomous is due lack interpretability AI making it challenging predict understand system's behavior under conditions.

Language: Английский

Photometric stereo multi-information fusion unsupervised anomaly detection algorithm

Applied Optics, Journal Year: 2024, Volume and Issue: 63(24), P. 6345 - 6345

Published: July 29, 2024

Due to different materials, product surfaces are susceptible light, shadow, reflection, and other factors. Coupled with the appearance of defects various shapes types, as well dust, impurities, interfering influences, normal abnormal samples difficult distinguish a common problem in field defect detection. Given this, this paper proposes an end-to-end photometric stereo multi-information fusion unsupervised anomaly detection model. First, feature generator is used obtain normal, reflectance, depth, information reconstruct 3D topographic details object’s surface. Second, multi-scale channel attention mechanism constructed fully use associations layers backbone network, limited enhance characterization ability. Finally, original image fused depth features find variability between defects, background. The differences source clone networks utilized achieve improve accuracy. In paper, model performance verified on PSAD dataset. experimental results show that algorithm has higher accuracy compared algorithms. Among them, input by 2.56% 1.57%, respectively. addition, ablation experiments further validate effectiveness paper.

Language: Английский

Developing Big Data Anomaly Dynamic and Static Detection Algorithms: AnomalyDSD Spark Package

Information Sciences, Journal Year: 2024, Volume and Issue: unknown, P. 121587 - 121587

Published: Oct. 1, 2024

Language: Английский