The performance comparison of support vector machine classification kernel functions on medical databases

In this research, an intelligent system framework was constructed by accurately comparing the classification performance of four different types of support vector machine; this included the SVM algorithm kernel functions (normalised polynomial kernel function (NPK), polynomial kernel function (PK), Pearson VII function-based Universal Kernel function (PUK), and the Radial Basis Function Kernel (RBF). This study used five different types of medical datasets (autistic children, autistic adolescents, chronic kidney failure, cryotherapy and immunotherapy), which differ from one another in terms of the quantity of the data and the medicinal and therapeutic content. The databases were extracted from the University of California Irvine machine learning repository. The method of tuning the parameters was followed in order to obtain the best performance results for the kernel functions using the Weka workbench tool. We then compared the best result of each kernel with the other kernels in terms of familiar classification standards in the field of data mining, which consisted of the confusion matrix, accuracy, sensitivity, precision and error rate.