Selection of Variables by the F-Score Algorithm for Radiated Magnetic Field Signals Discrimination of Electrical Discharges

This paper proposes the use of a hybrid variable selection method called F-Score with proposed threshold, which reduces the duration of the training and the testing phases while improving the accuracy of recognition for radiated magnetic field signals discrimination of electrical discharges generally taking place in insulation systems and in particular in insulators of high-voltage lines and power transformers. The classifier used is based on support vector machines (SVMs). The experimental analysis was carried out on a basis of data containing 161 signals which 106 signals were reserved for the training and 55 signals for the testing. The obtained results show that the proposed algorithm combined with the SVMs, allows a substantial reduction in the number of variables and a high improvement of the recognition rate compared to the pre-selection rate.