Performance Comparison of Support Vector Machine Kernel Functions in Classifying COVID-19 Sentiment

Abstract

Support Vector Machine (SVM) algorithm is a machine learning algorithm that is used to classify data by finding the best hyperplane that separates classes. In the SVM algorithm there are several types of kernel methods. Linear, Radial Basis Function (RBF), and polynomial kernel are some of the most commonly used SVM kernels. In previous research, each kernel has been used. However, the comparison of the three kernel function methods on the same dataset using accuracy, sensitivity, and specificity parameters has not been obtained. For this reason, this research is proposed to obtain comparative information of the three kernel functions using accuracy, sensitivity, and specificity parameters. The expected results can later be used as a reference for implementing the best kernel functions. The dataset used is comments on Youtube to analyze public sentiment on the increase in cases at the beginning of the entry of the COVID-19 pandemic in Indonesia. In this study, the accuracy values of the classification model were 0.86 for linear kernel, 0.90 for RBF kernel, and 0.91 for polynomial kernel. The sensitivity values obtained for each model are 0.64 for linear kernel, 0.48 for RBF kernel, and 0.20 for polynomial kernel. While the specificity values obtained for each model are 0.89 for linear kernel, 0.95 for RBF kernel, and 0.99 for polynomial kernel.