Enhancing Fault Diagnosis in Mechanical Systems with Graph Neural Networks Addressing Class Imbalance

Mathematics, Journal Year: 2024, Volume and Issue: 12(13), P. 2064 - 2064

Published: July 1, 2024

Recent advancements in intelligent diagnosis rely heavily on data-driven methods. However, these methods often encounter challenges adequately addressing class imbalances the context of fault mechanical systems. This paper proposes MeanRadius-SMOTE graph neural network (MRS-GNN), a novel framework designed to synthesize node representations GNNs effectively mitigate this issue. Through integrating oversampling technique into GNN architecture, MRS-GNN demonstrates an enhanced capability learn from under-represented classes while preserving intrinsic connectivity patterns data. Comprehensive testing various datasets superiority over traditional terms classification accuracy and handling imbalances. The experimental results three publicly available show that improves by 18 percentage points compared some popular Furthermore, exhibits higher robustness extreme imbalance scenarios, achieving AUC-ROC value 0.904 when rate is 0.4. not only enhances but also offers scalable solution applicable diverse complex systems, demonstrating its utility adaptability operating environments conditions.

Language: Английский

Improving Breast Cancer Detection Accuracy Through Random Forest Machine Learning Algorithm

Lecture notes in networks and systems, Journal Year: 2024, Volume and Issue: unknown, P. 318 - 338

Published: Jan. 1, 2024

Language: Английский