Parameter optimization algorithm of SVM for fault classification in traction converter

The classification performance of Support Vector Machine (SVM) is heavily influenced by its kernel parameter g and penalty factor c. in this paper, Cross-validation (CV) based grid-search optimization, CV-based genetic algorithm (GA) and CV-based particle swarm optimization (PSO) are respectively used for parameters optimization in SVM for fault classification of inverters in traction converter. Simulation result shows that SVM can reach the highest classification accuracy by using CV-based grid-search optimization algorithm, and it has been proved to be practical to use CV-based grid-search as SVM´s parameters optimization algorithm for fault classification in traction converter.