Analisis Komparatif IndoBERT, FinBERT, dan IndoLEM dengan Fine-Tuning untuk Klasifikasi Sentimen Berita Finansial Indonesia
DOI:
https://doi.org/10.30865/jurikom.v15i5.9867Keywords:
Analisis Sentimen, Berita Finansial, FinBERT, Fine-tuning, IndoBERT, IndoLEMAbstract
Analisis sentimen berita finansial memiliki peran penting dalam mendukung pengambilan keputusan investasi di pasar modal Indonesia. Penelitian terdahulu umumnya membandingkan satu model tanpa perbandingan lintas model, sehingga rekomendasi pemilihan model masih bersifat parsial. Penelitian ini membandingkan tiga model bahasa pre-trained, yaitu IndoBERT, IndoLEM, dan FinBERT, melalui pendekatan fine-tuning untuk klasifikasi sentimen tiga kelas (positif, netral, negatif) pada berita finansial berbahasa Indonesia. Dataset terdiri dari 9.015 artikel dari empat portal keuangan, dilabeli dengan strategi hybrid weak-supervision (lexicon-based labeling dan anotasi manusia). Ketiga model dilatih dengan konfigurasi hyperparameter identik menggunakan class-weighted cross-entropy loss. Evaluasi menggunakan accuracy, macro F1-score, weighted F1-score, dan Cohen’s Kappa pada validation set held-out (N = 2.149). Hasil eksperimen menunjukkan IndoBERT-Fine-Tuned mencapai Macro F1 tertinggi sebesar (0,863), diikuti IndoLEM-Fine-Tuned (0,843) dengan perbedaan prediksi antara kedua model yang signifikan berdasarkan uji McNemar (p = 0,0002), sementara FinBERT-Fine-Tuned mengalami penurunan tajam ke (0,437), bahkan di bawah baseline klasik TF-IDF + LinearSVC (0,809). Uji McNemar dengan koreksi Bonferroni mengonfirmasi bahwa kedua model BERT bahasa Indonesia secara signifikan mengungguli FinBERT-Fine-Tuned maupun baseline klasik (p < 0,001). Evaluasi pada gold subset berlabel manusia mengindikasikan adanya kemungkinan pengaruh pola label lexicon terhadap performa model pada label training. Kontribusi utama penelitian ini adalah evaluasi komparatif IndoBERT, IndoLEM, dan FinBERT pada berita finansial Indonesia dengan konfigurasi fine-tuning yang setara, disertai uji signifikansi statistik dan evaluasi pada gold standard berlabel manusia untuk menilai keterbatasan pendekatan weak-supervision.
References
[1] J. Hartanto, T. Liundi, R. Sutoyo, dan E. W. Andangsari, "ID-SMSA: Indonesian stock market dataset for sentiment analysis," Data in Brief, vol. 60, p. 111571, 2025, doi: 10.1016/j.dib.2025.111571.
[2] M. A. A. Saputra, "Deciphering news sentiment and stock price relationships in Indonesian companies: an AI-based exploration of industry affiliation and news co-occurrence," Discover Artificial Intelligence, vol. 5, no. 1, p. 87, 2025, doi: 10.1007/s44163-025-00350-5.
[3] A. H. Huang, H. Wang, dan Y. Yang, "FinBERT: A Large Language Model for Extracting Information from Financial Text," Contemporary Accounting Research, vol. 40, no. 2, pp. 806–841, 2023, doi: 10.1111/1911-3846.12832.
[4] F. Koto, A. Rahimi, J. H. Lau, dan T. Baldwin, "IndoLEM and IndoBERT: A Benchmark Dataset and Pre-trained Language Model for Indonesian NLP," dalam Proc. 28th Int. Conf. Computational Linguistics (COLING), Barcelona, 2020, pp. 757–770, doi: 10.18653/v1/2020.coling-main.66.
[5] E. Yulianti dan N. K. Nissa, "ABSA of Indonesian customer reviews using IndoBERT: single-sentence and sentence-pair classification approaches," Bulletin of Electrical Engineering and Informatics, vol. 13, no. 5, pp. 3579–3589, 2024, doi: 10.11591/eei.v13i5.8032.
[6] H. Imaduddin, F. Yusfida A’la, dan Y. S. Nugroho, "Sentiment Analysis in Indonesian Healthcare Applications using IndoBERT Approach," International Journal of Advanced Computer Science and Applications, vol. 14, no. 8, 2023, doi: 10.14569/IJACSA.2023.0140813.
[7] H. Ahmadian, T. F. Abidin, H. Riza, dan K. Muchtar, "Hybrid Models for Emotion Classification and Sentiment Analysis in Indonesian Language," Applied Computational Intelligence and Soft Computing, vol. 2024, art. 2826773, 2024, doi: 10.1155/2024/2826773.
[8] G. Z. Nabiilah, I. N. Alam, E. S. Purwanto, dan M. F. Hidayat, "Indonesian multilabel classification using IndoBERT embedding and MBERT classification," International Journal of Electrical and Computer Engineering, vol. 14, no. 1, pp. 1071–1078, 2024, doi: 10.11591/ijece.v14i1.pp1071-1078.
[9] J.-W. Kang dan S.-Y. Choi, "Comparative Investigation of GPT and FinBERT’s Sentiment Analysis Performance in News Across Different Sectors," Electronics, vol. 14, no. 6, p. 1090, 2025, doi: 10.3390/electronics14061090.
[10] N. Merayo, J. Vegas, C. Llamas, dan P. Fernández, "Social Network Sentiment Analysis Using Hybrid Deep Learning Models," Applied Sciences, vol. 13, no. 20, p. 11608, 2023, doi: 10.3390/app132011608.
[11] A. Khan, H. Zhang, N. Boudjellal, A. Ahmad, dan M. Khan, "Improving Sentiment Analysis in Election-Based Conversations on Twitter With ElecBERT Language Model," Computers, Materials & Continua, vol. 76, no. 3, pp. 3345–3361, 2023, doi: 10.32604/cmc.2023.041520.
[12] E. Dave dan A. Chowanda, "IPerFEX-2023: Indonesian personal financial entity extraction using indoBERT-BiGRU-CRF model," Journal of Big Data, vol. 11, no. 1, 2024, doi: 10.1186/s40537-024-00987-6.
[13] A. Kunaefi, Z. Abidin, dan R. Kusumawati, "Klasifikasi Berita Hoaks Bahasa Indonesia Menggunakan IndoBERT Fine-Tuning dengan Pendekatan Focal Loss pada Data Tidak Seimbang," JIPI (Jurnal Ilmiah Penelitian dan Pembelajaran Informatika), vol. 10, no. 2, pp. 1706–1714, 2025, doi: 10.29100/jipi.v10i2.7811.
[14] A. R. Hanum et al., "Analisis Kinerja Algoritma Klasifikasi Teks BERT Dalam Mendeteksi Berita Hoaks," Jurnal Teknologi Informasi dan Ilmu Komputer (JTIIK), vol. 11, no. 3, pp. 537–546, 2024, doi: 10.25126/jtiik938093.
[15] A. Jazuli, Widowati, dan R. Kusumaningrum, "Optimizing Aspect-Based Sentiment Analysis Using BERT for Comprehensive Analysis of Indonesian Student Feedback," Applied Sciences, vol. 15, no. 1, 2025, doi: 10.3390/app15010172.
[16] B. Wilie et al., "IndoNLU: Benchmark and Resources for Evaluating Indonesian Natural Language Understanding," dalam Proc. 1st Conf. Asia-Pacific Chapter of the ACL and 10th Int. Joint Conf. on NLP (AACL-IJCNLP), Suzhou, 2020, pp. 843–857, doi: 10.18653/v1/2020.aacl-main.85.
[17] L. Geni, E. Yulianti, dan D. I. Sensuse, "Sentiment Analysis of Tweets Before the 2024 Elections in Indonesia Using IndoBERT Language Models," Jurnal Ilmiah Teknik Elektro Komputer dan Informatika, vol. 9, no. 3, pp. 746–757, 2023, doi: 10.26555/jiteki.v9i3.26490.
[18] I. Daqiqil, H. Saputra, Syamsudhuha, R. Kurniawan, dan Y. Andriyani, "Sentiment analysis of student evaluation feedback using transformer-based language models," Indonesian Journal of Electrical Engineering and Computer Science, vol. 36, no. 2, pp. 1127–1139, 2024, doi: 10.11591/ijeecs.v36.i2.pp1127-1139.
[19] A. Areshey dan H. Mathkour, "Transfer Learning for Sentiment Classification Using Bidirectional Encoder Representations from Transformers (BERT) Model," Sensors, vol. 23, no. 11, p. 5232, 2023, doi: 10.3390/s23115232.
[20] V. D. Setiawan, D. U. Iswavigra, dan E. Anggiratih, "Implementation of IndoBERT for Sentiment Analysis of the Constitutional Court’s Decision Regarding the Minimum Age of Vice Presidential Candidates," Scientific Journal of Informatics, vol. 12, no. 3, pp. 397–406, 2025, doi: 10.15294/sji.v12i3.26360.
[21] A. Riyadi, M. Kovacs, U. Serdült, dan V. Kryssanov, "IndoGovBERT: A Domain-Specific Language Model for Processing Indonesian Government SDG Documents," Big Data and Cognitive Computing, vol. 8, no. 11, p. 153, 2024, doi: 10.3390/bdcc8110153.
[22] K. E. Saputra dan Riccosan, "Indonesian news article authorship attribution multilabel multiclass classification using IndoBERT," IAES International Journal of Artificial Intelligence, vol. 13, no. 4, pp. 4688–4694, 2024, doi: 10.11591/ijai.v13.i4.pp4688-4694.
[23] C. Shaw, P. LaCasse, dan L. Champagne, "Exploring emotion classification of Indonesian tweets using large scale transfer learning via IndoBERT," Social Network Analysis and Mining, vol. 15, no. 1, p. 22, 2025, doi: 10.1007/s13278-025-01439-6.
[24] S. Raza, D. Paulen-Patterson, dan C. Ding, "Fake news detection: comparative evaluation of BERT-like models and large language models with generative AI-annotated data," Knowledge and Information Systems, vol. 67, pp. 1–30, 2025, doi: 10.1007/s10115-024-02321-1.
