Siber Fiziksel Robot Kolunda Makine Öğrenmesi Bazlı Optimizasyon ve Anomali Tespiti Machine Learning Based Optimization and Anomaly Detection in a Cyber-Physical Robot Arm

Abstract

In this study, it is suggested to use Particle Swarm Optimization and Extra Trees methods together for the detection of anomalies that may occur due to various reasons during the operation of a cyber-physical robot arm. Particle Swarm Optimization (PSO) was used to determine which of the measured parameters are more useful for anomaly detection. PSO was preferred because it is a reliable method due to its very fast convergence and very low standard deviation. Extra Trees method was preferred for the detection of anomalies due to its high accuracy rate, being free from bias and low variation. With the combined use of these two methods, 99.16% accuracy, 98.65% precision, 98.68% sensitivity and 98.66% f1 score were achieved in the test data. This method gave more successful results in all performance criteria compared to the other methods in this study.