Comparison of Kernels on Support Vector Machine (SVM) Methods for Analysis of Cyberbullying

Along with the development of social media, there are various abuses and deviations from the ethics of interaction by its users. One form of social media abuse is cyberbullying. Recently, cyberbullying has become more frequent and these events are still difficult to detect. Therefore, researchers conducted an analysis of cyberbullying on Twitter social media using the Support Vector Machine (SVM). The data used is the crawled data of Twitter in Indonesian. This is included in non-linear data so it requires the kernel to carry out the text mining process. However, to date there has been no specific research on what kernels are good to use in cases of cyberbullying. Therefore, researchers conducted an experiment to implement SVM on tweet classification. The researchers will also look for the best kernels among the four kernels, namely Polynomial, Radial Basis Function (RBF), Sigmoid, and Linear. From the results of the experiments conducted, the SVM method can be used for the classification of cyberbullying. In addition, it is known that the sigmoid kernel has the highest accuracy of 83.85%.