Representasi Sentimen X (Twitter) Pilkada Jawa Timur dengan Hasil KPU Menggunakan Metode Transfer Learning IndoBERTweet

Ilham, Abdillah (2025) Representasi Sentimen X (Twitter) Pilkada Jawa Timur dengan Hasil KPU Menggunakan Metode Transfer Learning IndoBERTweet. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003211069-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Pemilihan Kepala Daerah (Pilkada) merupakan elemen penting dalam demokrasi di Indonesia, termasuk di Jawa Timur. Selama masa kampanye, media sosial, khususnya X (dulu dikenal sebagai Twitter), menjadi platform utama untuk menyuarakan opini politik. Analisis sentimen terhadap cuitan pengguna X dapat memberikan wawasan yang mendalam mengenai sentimen publik terhadap pasangan calon. Dalam penelitian ini, data cuitan dari X dikumpulkan melalui mekanisme crawling selama masa kampanye, yaitu dari 25 September 2024 hingga 23 November 2024. Pendekatan berbasis Transfer Learning menggunakan IndoBERTweet dipilih karena kemampuannya dalam memahami konteks bahasa Indonesia terlebih karena model ini dilatih pada korpus bahasa X. Untuk meningkatkan performa klasifikasi sentimen, IndoBERTweet dikombinasikan dengan arsitektur CNN, LSTM, dan Hybrid CNN-LSTM dengan penambahan class weight, augmentasi data back translation, dan regularisasi L2. Pemodelan terbaik didapatkan pada model IndoBERTweet Bi-LSTM yang mendapatkan akurasi dan weighted F1 Score pada data test sebesar 83% dan 81% tanpa indikasi overfitting tinggi. Hasil analisis sentimen, kemudian dianalisis menggunakan korelasi rank Kendall’s Tau dengan hasil rekapitulasi Komisi Pemilihan Umum (KPU). Hasilnya, persentase sentimen positif memiliki hubungan linear positif terhadap hasil rekapitulasi KPU. Hal ini berkebalikan dengan sentimen netral dan negatif yang memiliki hubungan linear negatif. Hasil sentimen juga dibandingkan dengan hasil survei dari waktu ke waktu dan ditemukan hubungan yang sedang–kuat, namun terkadang tidak konsisten dari waktu ke waktu dan untuk beberapa pasangan calon.
==================================================================================================================================
The Regional Head Election (Pilkada) is a crucial element of democracy in Indonesia, including in East Java. During the campaign period, social media, particularly X (formerly known as Twitter) emerged as a primary platform for expressing political opinions. Sentiment analysis of user tweets on X can offer deep insights into public sentiment toward the candidates. In this study, tweet data from X was collected through a crawling mechanism during the campaign period, from September 25, 2024, to November 23, 2024. A Transfer Learning approach using IndoBERTweet was selected due to its capability to understand the Indonesian language context, especially as the model was trained on X’s language corpus. To enhance sentiment classification performance, IndoBERTweet was combined with CNN, LSTM, and Hybrid CNN-LSTM architectures, incorporating class weights, back-translation data augmentation, and L2 regularization. The best model, IndoBERTweet Bi-LSTM achieved a test accuracy and weighted F1 score of 83% and 81%, respectively. Subsequently, the sentiment analysis results were examined using Kendall’s Tau rank correlation. The results indicate that the percentage of positive sentiment has a linear positive correlation with the KPU recapitulation results, while negative and neutral sentiments shows a linear negative relation. The sentiment result is also compared to survey result overtime and it is showed that it has a moderate to high positive relationship, although this relationship is not consistent overtime and for some candidate.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Analisis Sentimen, CNN, IndoBERTweet, Jatim, LSTM, Pilkada, CNN, IndoBERTweet, East Java, LSTM, Pilkada, Sentiment Analysis
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Abdillah Ilham
Date Deposited:	31 Jul 2025 03:00
Last Modified:	31 Jul 2025 03:00
URI:	http://repository.its.ac.id/id/eprint/123376

Actions (login required)

View Item