Selecting of the optimal feature subset and kernel parameters in digital modulation classification by using hybrid genetic algorithm–support vector machines: HGASVM

The support vector machines is a new technique for many pattern recognition areas. The digital modulation classification is one of these pattern recognition areas. In SVM training, the kernels, kernel parameters, and feature selection have very important roles for SVM classification accuracy. Therefore, most appropriates of these kernel types, kernel parameters and features should be used for SVM training. In this study, a hybrid of genetic algorithm–support vector machines (HGASVM) approach is presented in digital modulation classification area for increasing the support vector machines (SVM) classification accuracy. This HGASVM approach proposed in this paper selects of the optimal kernel function type, kernel function parameter, most appropriate wavelet filter type for problem, wavelet entropy parameter, and soft margin constant C penalty parameter of support vector machines (SVM) classifier. The classification accuracy of this HGASVM approach is tried by using real digital modulation dataset and compared with the SVMs, which has kernel function type, kernel function parameter, wavelet filter type, wavelet entropy parameter, and C parameter are randomly selected. Here, discrete wavelet transform (DWT) and adaptive wavelet entropy are used in feature extraction stage of this HGASVM approach. The digital modulation types used in this study are ASK-2, ASK-4, ASK-8, FSK-2, FSK-4, FSK-8, PSK-2, PSK-4, and PSK-8. The experimental studies conducted in this study show that the classification accuracy of this HGASVM approach is more superior than SVM, which has constant parameters.