Particle Swarm Optimization dan Genetic Algorithm untuk analisis sentimen pemekaran Papua di Twitter berbasis Support Vector Machine

Abstrak

Support Vector Machine (SVM) can classify sentiment analysis into positive or negative sentiment. In this study, sentiment data on the division of Papua was taken from Twitter. Because SVM has a weakness in feature selection during classification, the SVM algorithm optimization feature is implemented using feature selection, with Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). Tweets taken were 839 data tweets, then divided into 640 data for the training process and 199 data for the testing process. The data processing process is divided into two stages: data training and data testing. Tests were carried out in four models: the SVM, SVM+PSO, SVM+GA, SVM+PSO+GA algorithms. The experimental results show that SVM+PSO+GA modeling obtained the best accuracy results of 95.00%, with an AUC value of 0.912.