Danendra, Muhammad Rafi Widya (2026) Analisis Dinamika Sikap dan Polarisasi Opini Masyarakat untuk Memetakan Dinamika Wacana Politik di Media Sosial X menggunakan Pemodelan Topik serta Analisis Sikap dan Jaringan Sosial. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5026221088-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5026221088-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (14MB) | Request a copy |
Abstract
Media sosial X menjadi ruang penting bagi pembentukan dan pertukaran wacana politik, terutama pada tahun pertama pemerintahan Prabowo-Gibran. Akan tetapi, analisis wacana politik di media sosial sering masih dilakukan secara terpisah antara topik, sikap, dinamika temporal, dan struktur interaksi antarakun. Penelitian ini bertujuan mengintegrasikan analisis sikap, pemodelan dan klasifikasi topik, knowledge graph, serta analisis jaringan untuk memetakan dinamika wacana politik di media sosial X. Data dikumpulkan melalui strategi berbasis kata kunci, kemudian diproses melalui pra-pemrosesan teks, klasifikasi sikap menggunakan IndoBERT, penemuan awal topik menggunakan BERTopic, pelabelan manual, klasifikasi topik menggunakan IndoBERT, konstruksi knowledge graph di Neo4j, dan analisis komunitas dengan Louvain, PageRank, HTCI, serta stance balance. Hasil penelitian menunjukkan bahwa korpus final berjumlah 3.233 cuitan dengan delapan topik akhir. Model klasifikasi sikap memperoleh F1-macro sebesar 0,7018, sedangkan model klasifikasi topik memperoleh F1-macro sebesar 0,7243. Validasi anotasi eksternal menghasilkan Cohen's Kappa 0,8037 untuk sikap dan 0,7113 untuk topik sehingga mendukung konsistensi label dengan tetap memperhatikan catatan ambiguitas. Dari sisi substansi, Topik 6 terkait korupsi migas dan Topik 7 terkait kebijakan pangan cenderung lebih Pro-Pemerintah, sedangkan topik korupsi umum, evaluasi pemerintahan, dan kebijakan sosial-ekonomi lebih banyak memuat kritik. Knowledge graph menghasilkan 3.214 Tweet dan 3.615 User, dengan modularitas komunitas sebesar 0,9065 serta rasio interaksi internal antarkomunitas sebesar 93,84%. Temuan ini menunjukkan bahwa integrasi sikap, topik, waktu, dan relasi antarakun dapat memberi pembacaan wacana politik yang lebih mendalam. Akan tetapi, hasil penelitian tetap berlaku pada korpus berbasis kata kunci dan tidak dimaksudkan sebagai generalisasi penuh terhadap seluruh percakapan politik di X.
=======================================================================================================================================
Social media X has become an important space for the formation and exchange of political discourse, particularly during the first year of the Prabowo-Gibran administration. However, analyses of political discourse on social media are often conducted separately across topics, stance, temporal dynamics, and account-level interaction structures. This study aims to integrate stance analysis, topic modeling and classification, knowledge graph construction, and network analysis to map the dynamics of political discourse on social media X. Data were collected using a keyword-based strategy and processed through text preprocessing, IndoBERT-based stance classification, BERTopic-based topic discovery, manual labeling, IndoBERT-based topic classification, knowledge graph construction in Neo4j, and community analysis using Louvain, PageRank, HTCI, and stance balance. The results show that the final downstream corpus consists of 3,233 tweets grouped into eight final topics. The stance classification model achieved a macro-F1 score of 0.7018, while the topic classification model achieved a macro-F1 score of 0.7243. External annotation validation produced Cohen's Kappa scores of 0.8037 for stance and 0.7113 for topics, supporting label consistency while preserving caveats related to ambiguity. Substantively, Topic 6 on oil and gas corruption and Topic 7 on food policy tended to contain more pro-government stances, whereas topics related to general corruption, government evaluation, and socio-economic policy contained more critical stances. The knowledge graph produced a graph-main corpus consisting of 3,214 Tweet nodes and 3,615 User nodes, with a community modularity score of 0.9065 and an internal inter-community interaction ratio of 93.84%. These findings indicate that integrating stance, topics, time, and account-level relations provides a richer reading of political discourse. Nevertheless, the findings apply to the keyword-based corpus and are not intended as a full generalization of all political conversations on X.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Analisis Sikap, Pemodelan Topik, Analisis Jaringan Sosial, Polarisasi Opini, Media Sosial X, Stance Analysis, Topic Modeling, Knowledge Graph, Social Network Analysis, Opinion Polarization, Social Media X |
| Subjects: | H Social Sciences > HM Sociology > HM742 Online social networks. J Political Science > JA Political science (General) Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis |
| Depositing User: | Muhammad Rafi Widya Danendra |
| Date Deposited: | 28 Jul 2026 03:45 |
| Last Modified: | 28 Jul 2026 03:45 |
| URI: | http://repository.its.ac.id/id/eprint/138407 |
Actions (login required)
 |
View Item |