Prediction of Indonesia Presidential Election Results for the 2019-2024 Period Using Twitter Sentiment Analysis

The main purpose of this research is to predict the election of the President and Vice President of the Republic of Indonesia from 2019-2024 through the mining process of public opinion on twitter and test it accurately with classification algorithms namely Support Vector Machine (SVM) with selection features of Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). As the fourth largest democracy and the fifth largest twitter user in the world, twitter in Indonesia is very influential as a place for people to fight each other's arguments. Public talk about the election of two candidates for President and Vice President of the Republic of Indonesia for the 2019-2024 on Twitter, became an interesting topic for Twitter users with a variety of public sentiment both positive and negative. Therefore, there is a need for a method that helps to see public opinion effectively. The researchers used tweets in Indonesian as research data with keywords #jokowi2periode and #2019tetapjokowi, and #2019prabowosandi and #2019gantipresiden with a total of 4000 tweets. This study uses a classification technique, namely SVM Algorithm. This method is chosen based on many of the best classification techniques commonly used for the analysis of opinion sentiments. It is because of SVM has weaknesses for selecting the appropriate parameters or features. In this study, researchers made improvements to previous research using a feature selection comparison method, PSO & GA. The results of this study are in the form of a prediction of the pairs of candidates for the President and the Vice President of Indonesia for the period of 2019-2024 who have more positive sentiments. Based on public opinion on Twitter, the pair Prabowo Subianto-Sandiaga Uno is predicted to be elected as President and Vice President of Indonesia for the period of 2019-2024 with the most positive sentiment, reaching 830 out of 1000 tweets entered. And the SVM method of the combination of PSO is the best method with accuracy reaching 86.20% and the AUC value reaching 0.934.