Machine Learning Algorithms and Fault Detection for Improved Belief Function Based Decision Fusion in Wireless Sensor...

Area of Science:

• Computer Science
• Electrical Engineering
• Network Engineering

Background:

• Decision fusion enhances classification accuracy and reduces data transmission costs in Wireless Sensor Networks (WSNs).
• Decentralized classification fusion in WSNs necessitates belief function-based approaches.
• WSNs are susceptible to faults due to hardware/software issues and environmental factors, requiring efficient fault detection.

Purpose of the Study:

• To improve belief function-based decision fusion in WSNs.
• To propose and evaluate four enhanced classification techniques: EKNN, EELM, ESVM, and ERELM.
• To address sensor failure issues in WSNs through enhanced classification methods for fault detection.

Main Methods:

• Proposed four enhanced classification techniques: Enhanced K-Nearest Neighbor (EKNN), Enhanced Extreme Learning Machine (EELM), Enhanced Support Vector Machine (ESVM), and Enhanced Recurrent Extreme Learning Machine (ERELM).
• Induced four types of sensor faults: offset, gain, stuck-at, and out of bounds.
• Evaluated fault detection performance using Detection Accuracy (DA), True Positive Rate (TPR), and Error Rate (ER).

Main Results:

• ERELM achieved the best performance in improving belief function fusion.
• ESVM, EELM, and EKNN provided the second, third, and fourth best results, respectively.
• The proposed enhanced classifiers outperformed existing techniques in belief function fusion and fault detection.

Conclusions:

• The proposed enhanced classification methods effectively improve belief function fusion and fault detection in WSNs.
• ERELM is the most effective among the proposed methods for these tasks.
• The study highlights the potential of enhanced classifiers for robust WSN operation despite sensor failures.