Apriansyah, Muhammad Rifqi and Kusuma, Dine Tiara (2026) Komparasi Latent Dirichlet Allocation dan BERTopic Modeling dalam Identifikasi Topik pada ulasan KPU di X. Diploma thesis, Institut Teknologi PLN.
Cover.pdf
Download (121kB)
Lembar Pengesahan.pdf
Download (294kB)
Abstrak.pdf
Download (173kB)
Bab 1.pdf
Download (164kB)
Bab 2.pdf
Download (488kB)
Bab 3.pdf
Restricted to Registered users only
Download (1MB)
Bab 4.pdf
Restricted to Registered users only
Download (1MB)
Bab 5.pdf
Restricted to Registered users only
Download (127kB)
Daftar Pustaka.pdf
Download (168kB)
File Secara Utuh.pdf
Restricted to Registered users only
Download (2MB)
Abstract
Media sosial X menjadi sarana utama masyarakat dalam menyampaikan opini
terhadap kinerja Komisi Pemilihan Umum (KPU), khususnya pada konteks Pemilu 2024.
Banyaknya data teks yang bersifat tidak terstruktur menuntut penggunaan metode
pemodelan topik untuk mengidentifikasi isu-isu utama yang berkembang. Penelitian ini
bertujuan untuk mengidentifikasi topik percakapan publik mengenai KPU di X serta
membandingkan kinerja metode Latent Dirichlet Allocation (LDA) dan BERTopic dalam
pemodelan topik. Metode penelitian meliputi pengumpulan data komentar terkait KPU,
tahap prapemrosesan teks, pemodelan topik menggunakan LDA dan BERTopic, serta
evaluasi kualitas topik menggunakan nilai koherensi dan interpretabilitas. Hasil penelitian
menunjukkan bahwa BERTopic menghasilkan topik yang lebih koheren dan kontekstual
dibandingkan LDA karena memanfaatkan representasi semantik berbasis embedding
transformer. Sementara itu, LDA mampu memodelkan distribusi kata secara probabilistik
namun cenderung menghasilkan topik yang lebih umum pada teks pendek. Dengan
demikian, BERTopic lebih efektif dalam mengidentifikasi topik berbasis makna pada
ulasan KPU di X, sedangkan LDA lebih sesuai untuk analisis distribusi kata. Penelitian
ini diharapkan dapat memberikan gambaran isu publik dominan serta menjadi referensi
pemilihan metode pemodelan topik pada data media sosial.
Social media platform X has become a major medium for the public to express
opinions regarding the performance of the General Election Commission (KPU),
particularly in the context of the 2024 General Election. The large volume of unstructured
textual data requires the application of Topic Modeling methods to identify the main
issues discussed by the public. This study aims to identify dominant topics in public
discussions about KPU on X and to compare the performance of Latent Dirichlet
Allocation (LDA) and BERTopic in Topic Modeling. The research methodology includes
data collection of KPU-related comments, text preprocessing, Topic Modeling using LDA
and BERTopic, and evaluation of topic quality based on coherence value and
interpretability. The results show that BERTopic produces more coherent and contextually
meaningful topics than LDA by utilizing transformer-based semantic embeddings.
Meanwhile, LDA is effective in modeling word distributions probabilistically but tends to
generate more general topics when applied to short texts. Therefore, BERTopic is more
suitable for semantic-based topic identification in KPU-related reviews on X, while LDA
is more appropriate for word distribution analysis. This study is expected to provide
insights into dominant public issues and serve as a reference for selecting Topic Modeling
methods for social media text analysis.
| Item Type: | Thesis (Diploma) |
|---|---|
| Uncontrolled Keywords: | Topic Modeling, Latent Dirichlet Allocation, BERTopic, Media Sosial X, KPU. Topic Modeling, Latent Dirichlet Allocation, BERTopic, Social Media X, KPU |
| Subjects: | Skripsi Bidang Keilmuan > Teknik Informatika |
| Divisions: | Fakultas Telematika Energi > S1 Teknik Informatika |
| Depositing User: | Mr Apriansyah Muhammad Rifqi |
| Date Deposited: | 04 Mar 2026 06:31 |
| Last Modified: | 04 Mar 2026 06:31 |
| URI: | https://repository.itpln.ac.id/id/eprint/5660 |
