Exploiting Linear Support Vector Machine for Correlation-Based High Dimensional Data Classification in Wireless...

Area of Science:

• Computer Science
• Machine Learning
• Wireless Sensor Networks

Background:

• Linear Support Vector Machine (LSVM) is effective for link classification in sensor networks.
• Kernelized Linear Support Vector Machine (KLSVM) offers stable and consistent results by learning optimal parameter settings.
• Wireless Sensor Networks (WSNs) present challenges in high-dimensional classification and spatiotemporal data correlation.

Purpose of the Study:

• To present a data-driven framework for reliable link classification using KLSVM.
• To investigate KLSVM's application in modeling high-dimensional multi-category classification and spatiotemporal data correlation in WSNs.
• To evaluate KLSVM's effectiveness in anomaly detection within WSNs.

Main Methods:

• Utilized Kernelized Linear Support Vector Machine (KLSVM) for link classification.
• Optimized linear kernel hyperparameters for high-dimensional data classification.
• Analyzed packet traces from an 802.15.4 network WSN testbed.

Main Results:

• KLSVM successfully modeled high-dimensional data classification and spatiotemporal correlations.
• Packet Reception Rate (PRR) > 50% showed high negative correlation; other observations showed moderate positive correlation.
• The model provided visual insights into network behavior and demonstrated good link quality estimation accuracy.

Conclusions:

• KLSVM provides a robust framework for reliable link classification in WSNs.
• The technique accurately detects anomalies and estimates link quality in sensor networks.
• KLSVM offers a data-driven approach for understanding and managing WSNs effectively.