ANALISIS SENTIMEN DAN DETEKSI BUZZER TERHADAP PASLON GUBERNUR DALAM KAMPANYE PILKADA JAKARTA 2024 DENGAN NAÏVE BAYES DAN GAUSSIAN NAÏVE BAYES

SYALVA SYADILA, . (2025) ANALISIS SENTIMEN DAN DETEKSI BUZZER TERHADAP PASLON GUBERNUR DALAM KAMPANYE PILKADA JAKARTA 2024 DENGAN NAÏVE BAYES DAN GAUSSIAN NAÏVE BAYES. Sarjana thesis, UNIVERSITAS NEGERI JAKARTA.

	Text Cover.pdf Download (1MB)
	Text BAB I.pdf Download (320kB)
	Text BAB II.pdf Restricted to Registered users only Download (577kB) | Request a copy
	Text BAB III.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB IV.pdf Restricted to Registered users only Download (1MB) | Request a copy
	Text BAB V.pdf Restricted to Registered users only Download (395kB) | Request a copy
	Text Daftar Pustaka.pdf Download (269kB)
	Text Lampiran.pdf Restricted to Registered users only Download (2MB) | Request a copy

Abstract

Media sosial X (sebelumnya Twitter) menjadi ruang perbincangan politik selama Pilkada Jakarta 2024, dengan 24,69 juta pengguna aktif di Indonesia. Sifatnya yang real-time menjadikan platform ini strategis untuk kampanye, namun juga rawan dimanfaatkan oleh buzzer untuk membentuk persepsi publik secara tidak organik, sehingga opini yang tersebar tidak mencerminkan suara masyarakat sebenarnya. Penelitian ini menyajikan pendekatan untuk menganalisis sentimen terhadap tiga pasangan calon gubernur serta mendeteksi buzzer untuk mengevaluasi pengaruh buzzer dalam membentuk persepsi publik selama Pilkada Jakarta 2024. Pertama, data dikumpulkan melalui crawling selama masa kampanye (25 September - 23 November 2024), dan setelah dilakukan pre-processing menghasilkan 7.781 tweet. Kedua, deteksi buzzer menggunakan algoritma Gaussian Naïve Bayes berdasarkan fitur seperti followers, following, age, tweet per day, active weeks count, dan positive sentiment ratio. Ketiga, analisis sentimen dilakukan dengan Multinomial Naïve Bayes untuk menentukan polaritas tiap tweet. Terakhir, distribusi sentimen dibandingkan sebelum dan sesudah buzzer disaring. Hasil menunjukkan bahwa 55% tweet berasal dari 25 akun buzzer. Model deteksi buzzer mencapai akurasi 90%, dan model sentimen 85,65%. Buzzer paling banyak memengaruhi Paslon 1 (58,3%), diikuti Paslon 2 (57,6%) dan Paslon 3 (39%). Aktivitas buzzer memuncak pada 14, 16, dan 17 November 2024, bertepatan dengan momen pra dan pasca debat. Rata-rata tweet buzzer dalam 3 hari tersebut mencapai 184 tweet. Standar deviasi aktivitas buzzer sebesar 22,5, jauh lebih tinggi dibanding non-buzzer sebesar 0,32. Hal ini menunjukkan bahwa buzzer beroperasi secara terorganisir dan masif, serta lebih banyak menyebarkan sentimen positif dalam strategi kampanye bertipe positive astroturfing. ******* Social media platform X (formerly Twitter) served as a major arena for political discourse during the 2024 Jakarta gubernatorial election, with 24.69 million active users in Indonesia. Its real-time and open nature makes it a strategic tool for political campaigns, but also vulnerable to manipulation by buzzers who shape public perception inorganically, thus potentially distorting the actual voice of the people. This study presents an approach to analyze sentiment toward three gubernatorial candidates and detect buzzer activity to evaluate its influence on public opinion throughout the campaign period. First, data were collected through crawling during the campaign period (25 September - 23 November 2024), resulting in 7,782 tweets after pre-processing. Second, buzzer detection was performed using the Gaussian Naïve Bayes algorithm based on features such as followers, following, account age, tweet per day, active weeks count, and positive sentiment ratio. Third, sentiment analysis was carried out using the Multinomial Naïve Bayes algorithm to determine the polarity of each tweet. Lastly, sentiment distribution was compared before and after buzzer filtering. Results show that 55% of the tweets originated from 25 buzzer accounts. The buzzer detection model achieved 90% accuracy, while the sentiment model reached 85.65%. Buzzers had the strongest influence on Candidate 1 (58.3%), followed by Candidate 2 (57.6%) and Candidate 3 (39%). Their activity peaked on 14, 16, and 17 November 2024—coinciding with pre- and post-debate periods. Buzzers averaged 184 tweets during these three days, with a daily activity standard deviation of 22.5, significantly higher than non-buzzers at 0.32. These findings indicate that buzzers operate in an organized and large-scale manner, predominantly spreading positive sentiment through a positive astroturfing strategy.

Item Type:	Thesis (Sarjana)
Additional Information:	1). Lipur Sugiyanta, Ph.D. ; 2). Murien Nugraheni, M.Cs.
Subjects:	Teknologi dan Ilmu Terapan > Teknik Komputer
Divisions:	FT > S1 Sistem dan Teknologi Informasi
Depositing User:	Users 27882 not found.
Date Deposited:	24 Jul 2025 01:31
Last Modified:	24 Jul 2025 01:31
URI:	http://repository.unj.ac.id/id/eprint/56678

Actions (login required)

View Item