PARTICLE SWARM OPTIMIZATION IN TWIN SUPPORT VECTOR MACHINE TO CLASSIFY FAKE NEWS

Abstract

Basically online media give freedom to the user to speak with other users. Because of this freedom, some users misuse online media, namely by spreading information that cannot be justified or what is called fake news. To avoid the spread of fake news, users must know whether the news is fake news or not. In this study, the classification of news will be carried out. This research use Twin Support Vector Machine (TWSVM) as classification method. However TWSVM has a disadvantage viz difficult to determine optimal parameter and that's why an optimization is needed to find the parameter with Particle Swarm Optimization (PSO). There are five main parameters used in this research viz í µí° ¶₁ and í µí° ¶₂ , w, number of particle and iteration number. Model performance optimization will be seen from impact of the parameters. Model performance will be measured with evaluation matrix accuracy, precision, recall, and F1-Score. The result of this research show that Particle Swarm Optimization can improve the accuracy, precision, and f-1 score of classification model Twin Support Vector Machine.