Electronic Theses and Dissertation

Pemilihan presiden di Indonesia pada tahun 2024 menunjukkan peran signifikan media sosial X dalam membentuk opini publik. Penelitian ini bertujuan untuk menganalisis perubahan opini publik dengan mengeksplorasi fluktuasi sentimen terhadap tiga pasangan calon presiden selama kampanye, serta mengidentifikasi topik utama dan hashtag populer pada periode anomali menggunakan pendekatan BERTopic. Dataset penelitian terdiri dari 9.817.355 baris yang mencakup interaksi di media sosial X. Analisis sentimen dilakukan menggunakan model Indonesia-RoBERTa, sedangkan topic modelling diterapkan dengan BERTopic untuk mengidentifikasi topik utama dalam percakapan. Selain itu, Social Network Analysis (SNA) digunakan untuk memetakan pola hubungan antar hashtag, dengan pengukuran degree centrality, closeness centrality, dan eigenvector centrality guna menilai keterhubungan dan dominasi hashtag dalam jaringan diskusi. Hasil penelitian menunjukkan bahwa: 44,47% tweet bersentimen netral; 29,23% negatif; dan 26,10% positif. Lonjakan sentimen negatif yang signifikan terjadi pada 21 Januari 2024, bertepatan dengan debat calon wakil presiden keempat. Hasil topic modelling menunjukkan kontroversi program food estate, kritik terhadap kinerja politikus, dan polarisasi pendukung pasangan calon. Visualisasi jaringan hashtag menunjukkan dominasi #GanjarMahfud2024 yang memiliki degree centrality tertinggi sebesar 562, closeness centrality sebesar 0,5, dan eigenvector centrality dengan nilai sempurna 1,0. Hal ini mencerminkan tingginya intensitas diskusi dan mobilisasi dukungan di media sosial. Hasil penelitian ini dapat menjadi acuan bagi tim kampanye dalam merancang strategi komunikasi yang lebih efektif, bagi media dalam menyajikan analisis mendalam, serta bagi masyarakat dalam memahami isu-isu utama selama pemilu yang dapat memengaruhi preferensi pemilih.

The presidential election in Indonesia in 2024 demonstrated the significant role of X social media in shaping public opinion. This study aims to analyze changes in public opinion by exploring sentiment fluctuations towards three presidential candidate pairs during the campaign, as well as identifying key topics and popular hashtags in anomalous periods using the BERTopic approach. The research dataset consists of 9,817,355 rows covering interactions on X social media. Sentiment analysis was conducted using the Indonesia-RoBERTa model, while topic modeling was applied with BERTopic to identify the main topics in the conversation. In addition, Social Network Analysis (SNA) was used to map the relationship patterns between hashtags, with degree centrality, closeness centrality and eigenvector centrality measurements to assess the connectedness and dominance of hashtags in the discussion network. The results showed that: 44.47% of tweets had neutral sentiment; 29.23% were negative; and 26.10% were positive. A significant spike in negative sentiment occurred on January 21, 2024, coinciding with the fourth vice presidential candidate debate. The results of topic modeling showed controversy over the food estate program, criticism of politicians' performance, and polarization of supporters of candidate pairs. The hashtag network visualization shows the dominance of #GanjarMahfud2024 which has the highest degree centrality of 562, closeness centrality of 0.5, and eigenvector centrality with a perfect value of 1.0. This reflects the high intensity of discussion and mobilization of support on social media. The results of this study can be a reference for campaign teams in designing more effective communication strategies, for the media in presenting in-depth analysis, and for the public in understanding the main issues during the election that can influence voter preferences.