SISTEM GovBERTic UNTUK ANALISIS OPINI PUBLIK TERHADAP KINERJA SATU TAHUN PERTAMA PRABOWO-GIBRAN (STUDI KASUS: MEDIA SOSIAL X)

Nariswari, Naura Ulayya (2026) SISTEM GovBERTic UNTUK ANALISIS OPINI PUBLIK TERHADAP KINERJA SATU TAHUN PERTAMA PRABOWO-GIBRAN (STUDI KASUS: MEDIA SOSIAL X). Undergraduate thesis, UPN "VETERAN" JAWA TIMUR.

	Text (cover) 22083010034-cover (2).pdf Download (978kB)
	Text (bab 1) 22083010034-bab1.pdf Download (229kB)
	Text (bab 2) 22083010034-bab2.pdf Restricted to Repository staff only until 22 July 2029. Download (776kB)
	Text (bab 3) 22083010034-bab3.pdf Restricted to Repository staff only until 22 July 2029. Download (555kB)
	Text (bab 4) 22083010034-bab4.pdf Restricted to Repository staff only until 22 July 2029. Download (922kB)
	Text (bab 5) 22083010034-bab5.pdf Download (203kB)
	Text (daftar pustaka) 22083010034-daftarpustaka.pdf Download (182kB)
	Text (lampiran) 22083010034-lampiran.pdf Restricted to Repository staff only Download (224kB)

Abstract

In the first year of the new President and Vice President's administration, social media became the main space for the public to express their opinions on government policies and performance. However, the narrative that develops on Platform X is often dominated by claims that are not supported by transparent analysis methodologies that have the potential to shape biased public opinion. This research aims to build a GovBERTic system that integrates topic modeling and sentiment analysis to identify key issues as well as the tendency of public sentiment towards government performance based on data from Platform X. The topic modeling stage was carried out using BERTopic with GovBERTic-based semantic representation, dimension reduction using UMAP, clustering using HDBSCAN, and keyword extraction through c-TF-IDF and KeyBERT. Sentiment analysis was carried out by comparing three transformer-based models , namely IndoBERT, IndoBERTweet, and RoBERTa, using the Stratified 5-Fold Cross Validation method with accuracy, precision, recall, and F1-score metrics. The research dataset consisted of 5,787 tweets with 5,538 clean tweets to analyze. The results of the topic modeling showed that BERTopic managed to produce eight main topics with a Topic Diversity of 0.9375 and a Coherence Score (Cv) value of 0.4326 which shows good topic quality on social media data. At the sentiment analysis stage, RoBERTa performed better than IndoBERT and IndoBERTweet with an average accuracy of 74.44% ± 9.69% and a macro F1-score of 73.75% ± 10.57%, so it was chosen as the sentiment classification model. The integration of topic modeling and sentiment analysis results showed that of the eight topics formed, there were four topics with neutral dominant sentiment, two topics with negative dominant sentiment, and two topics with positive dominant sentiment. The results of the study show that GovBERTic is able to identify the main issues and trends of public sentiment objectively and based on data, so that it can be an analytical instrument that supports the evaluation of public perception of government performance on social media. Keywords: sentiment analysis, BERTopic, GovBERTic, social media X, Prabowo–Gibran government.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Muhaimin, Amri	NIDN2119950723270	amri.muhaimin.stat@upnjatim.ac.id
Thesis advisor	Maywara, Shindi Shella	NUPTK1850774675230252	shindi.shella.fasilkom@upnjatim.ac.id

Subjects:

H Social Sciences > HG Finance > HG1709 Data processing
J Political Science > JA Political science (General)

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Ms Naura Nariswari

Date Deposited:

22 Jul 2026 08:02

Last Modified:

23 Jul 2026 01:23

URI:

https://repository.upnjatim.ac.id/id/eprint/57605