A ANALISIS SENTIMEN PARA KANDIDAT PILPRES 2024 DENGAN MODEL BAHASA BERT

Abstract

Indonesia has conducted a general election for presidential and legislative positions for central and regional government on 14 February 2024. With the advancement of technology, each candidate has utilized social media to engage with the voters. One of the most common platforms for people to have political discussions is X. This study aims to perform sentiment analysis better to understand people's perceptions towards each 2024 presidential candidate. Using BERT as the primary model, this research classifies sentiment toward each candidate separately with only two labels: positive and negative. The metrics used in this study are accuracy and F1-score with astonishing results, 85% and 89% for Anies’s dataset, 88% and 89% for Prabowo’s dataset, and 89% for both metrics for Ganjar’s dataset. This research also generated keywords for each sentiment towards candidates from BERT’s validation data predictions. This study contributes insights into the narratives and sentiments surrounding the 2024 presidential election.