IPB University Logo

SCIENTIFIC REPOSITORY

IPB University Scientific Repository collects, disseminates, and provides persistent and reliable access to the research and scholarship of faculty, staff, and students at IPB University

AI Repository
 
Building and Categories


      View Item 
      •   IPB Repository
      • Final Assignments
      • Undergraduate Final Assignments
      • UF - School of Data Science, Mathematic and Informatics
      • UF - Statistics and Data Sciences
      • View Item
      •   IPB Repository
      • Final Assignments
      • Undergraduate Final Assignments
      • UF - School of Data Science, Mathematic and Informatics
      • UF - Statistics and Data Sciences
      • View Item
      JavaScript is disabled for your browser. Some features of this site may not work without it.

      Analisis Sentimen Publik terhadap Kebijakan Penempatan Dana Negara di Bank Umum Mitra Menggunakan IndoBERT-SupCon

      Thumbnail
      View/Open
      Cover (469.1Kb)
      Fulltext (7.143Mb)
      Lampiran (229.6Kb)
      Date
      2026
      Author
      Af'ida, Biki Nurul
      Angraini, Yenni
      Firdawanti, Aulia Rizki
      Metadata
      Show full item record
      Abstract
      Pemerintah Indonesia menetapkan kebijakan penempatan dana negara sebesar Rp200 triliun pada bank umum mitra melalui Keputusan Menteri Keuangan Nomor 276 Tahun 2025. Kebijakan tersebut memunculkan beragam respons publik di media sosial, termasuk YouTube sebagai ruang diskusi terbuka. Karakteristik komentar YouTube yang singkat, informal, dan ambigu menyulitkan proses klasifikasi sentimen. Oleh karena itu, penelitian ini menerapkan IndoBERT-SupCon, yaitu pengembangan IndoBERT dengan supervised contrastive learning untuk meningkatkan kemampuan model dalam membedakan representasi antarkelas sentimen. Penelitian ini bertujuan mengidentifikasi kecenderungan sentimen publik terhadap kebijakan tersebut serta mengevaluasi performa IndoBERT-SupCon dalam klasifikasi sentimen komentar YouTube. Data penelitian berupa komentar YouTube yang dikumpulkan menggunakan YouTube Data API v3. Hasil pelabelan final menunjukkan bahwa sentimen publik didominasi sentimen negatif sebanyak 4592 komentar (60,5%), diikuti sentimen positif sebanyak 2205 komentar (29,1%), dan sentimen netral sebanyak 792 komentar (10,4%). IndoBERT-SupCon dengan ?=0,5 dan t=0,07 memperoleh macro F1-score sebesar 0,7095±0,0092, lebih tinggi dibandingkan IndoBERT baseline sebesar 0,7008±0,0085. Hasil tersebut menunjukkan bahwa penambahan supervised contrastive learning memberikan peningkatan performa pada klasifikasi sentimen komentar YouTube, meskipun masih relatif terbatas karena IndoBERT baseline sudah memiliki kemampuan representasi yang kuat, komentar YouTube bersifat informal dan ambigu, serta efektivitas SupCon dapat dipengaruhi oleh pasangan positif dan negatif dalam batch.
       
      The Indonesian government implemented a policy to allocate IDR 200 trillion of state funds to partner commercial banks through Minister of Finance Decree Number 276 of 2025. This policy generated diverse public reactions on social media, including YouTube as an open discussion platform. The short, informal, and ambiguous characteristics of YouTube comments make sentiment classification challenging. Therefore, this study employed IndoBERT-SupCon, an extension of IndoBERT with supervised contrastive learning to improve the model’s ability to distinguish representations across sentiment classes. This study aimed to identify public sentiment toward the policy and evaluate the performance of IndoBERT-SupCon in classifying YouTube comments. The dataset consisted of YouTube comments collected using the YouTube Data API v3. The final labeling results showed that public sentiment was dominated by negative sentiment, with 4592 comments (60.5%), followed by 2205 positive comments (29.1%), and 792 neutral comments (10.4%). IndoBERT-SupCon with ?=0.5 and t=0.07 achieved a macro F1-score of 0.7095±0.0092, outperforming the IndoBERT baseline (0.7008±0.0085). The results indicated that supervised contrastive learning improved sentiment classification, although the improvement remained limited because the IndoBERT baseline already had strong representation capability, YouTube comments were generally informal and ambiguous, and SupCon effectiveness depended on positive and negative pairs within each batch.
       
      URI
      http://repository.ipb.ac.id/handle/123456789/175652
      Collections
      • UF - Statistics and Data Sciences [166]

      Copyright © 2020 Library of IPB University
      All rights reserved
      Contact Us | Send Feedback
      Indonesia DSpace Group 
      IPB University Scientific Repository
      UIN Syarif Hidayatullah Institutional Repository
      Universitas Jember Digital Repository
        

       

      Browse

      All of IPB RepositoryCollectionsBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

      My Account

      Login

      Application

      google store

      Copyright © 2020 Library of IPB University
      All rights reserved
      Contact Us | Send Feedback
      Indonesia DSpace Group 
      IPB University Scientific Repository
      UIN Syarif Hidayatullah Institutional Repository
      Universitas Jember Digital Repository