Analisis Sentimen Calon Presiden 2024 di Media Sosial X Menggunakan Naive Bayes dan SMOTE

Abstract

In the era of digital advancement, the utilization of social media has surged, enabling individuals to express their viewpoints openly. This research underscores the utilization of social media platform X as the primary avenue for users to express their opinions, particularly on political matters, notably within the framework of the presidential election. Sentiment analysis techniques, specifically employing the Naïve Bayes algorithm and the Synthetic Minority Oversampling (SMOTE) method, have been the central focus of inquiry to infer people's inclinations toward presidential candidates. Despite numerous antecedent studies, deficiencies persist in terms of precision and data imbalance. This study endeavors to enhance the efficacy of sentiment analysis by integrating the Naïve Bayes approach with SMOTE. By scrutinizing tweets on social media X spanning from December 12, 2023, to March 31, 2024, the data is categorized into positive and negative sentiments. The findings reveal that employing SMOTE bolstered accuracy to 88% for the Ganjar-Mahfud dataset, whereas accuracy without SMOTE languished at approximately 69% for the Anies-Imin dataset. Out of 1589 tweets conveying positive sentiments, approximately 27.7% were directed towards Anies-Imin, 28.7% towards Prabowo-Gibran, and 43.5% towards Ganjar-Mahfud. The preponderance of negative sentiments was aimed at Anies-Imin (41.5%) and Prabowo-Gibran (40.8%).