PENGEMBANGAN SISTEM PERINGKASAN OTOMATIS BERBASIS LARGE LANGUAGE MODELS UNTUK DOKUMEN LAPORAN TAHUNAN PERUSAHAAN TERBUKA

    Muhammad Rizki, - and Yudi Wibisono, - and Eddy Prasetyo Nugroho, - (2025) PENGEMBANGAN SISTEM PERINGKASAN OTOMATIS BERBASIS LARGE LANGUAGE MODELS UNTUK DOKUMEN LAPORAN TAHUNAN PERUSAHAAN TERBUKA. S1 thesis, Universitas Pendidikan Indonesia.

    Abstract

    Penelitian ini bertujuan untuk mengembangkan sistem peringkasan otomatis laporan tahunan perusahaan yang terdaftar di Bursa Efek Indonesia (BEI) guna mendukung proses analisis fundamental oleh investor. Dataset peringkasan disusun dari bagian-bagian naratif laporan tahunan yang telah diekstraksi dan dianotasi, kemudian digunakan untuk melakukan fine-tuning terhadap Large Language Model (LLM) LLaMA 3.2-3B dengan metode Low-Rank Adaptation (LoRA). Evaluasi kuantitatif menggunakan metrik ROUGE dan BERTScore menunjukkan bahwa model hasil fine-tuning mengalami peningkatan performa signifikan dibandingkan model dasar, dengan skor ROUGE-1 sebesar 0.5624, ROUGE-2 sebesar 0.2714, ROUGE-L sebesar 0.3841, dan BERTScore F1 sebesar 0.9105. Selain itu, model ini juga melampaui performa model generalis dengan ukuran yang lebih besar, Mistral 7B. Luaran akhir penelitian adalah aplikasi berbasis web yang mampu menerima dokumen laporan tahunan dalam format PDF dan menghasilkan ringkasan naratif secara otomatis. This study aims to develop an automatic summarization system for the annual reports of companies listed on the Indonesia Stock Exchange (IDX) to support investors in conducting fundamental analysis. The summarization dataset was compiled from the narrative sections of annual reports that had been extracted and annotated, and subsequently used to fine-tune the LLaMA 3.2-3B Large Language Model (LLM) using the Low-Rank Adaptation (LoRA) method. Quantitative evaluation using ROUGE and BERTScore metrics shows that the fine-tuned model achieved a significant performance improvement over its base model, attaining ROUGE-1 of 0.5624, ROUGE-2 of 0.2714, ROUGE-L of 0.3841, and BERTScore F1 of 0.9105. Furthermore, the model outperformed a larger generalist model, Mistral 7B. The final outcome of this research is a web-based application capable of accepting annual report documents in PDF format and automatically generating narrative summaries.

    Baca Full Text klik disini
    [thumbnail of S_KOM_2107922_Title.pdf] Text
    S_KOM_2107922_Title.pdf
    Download (646kB)
    [thumbnail of S_KOM_2107922_Chapter1.pdf] Text
    S_KOM_2107922_Chapter1.pdf
    Download (279kB)
    [thumbnail of S_KOM_2107922_Chapter2.pdf] Text
    S_KOM_2107922_Chapter2.pdf
    Restricted to Staf Perpustakaan
    Download (1MB)
    [thumbnail of S_KOM_2107922_Chapter3.pdf] Text
    S_KOM_2107922_Chapter3.pdf
    Download (365kB)
    [thumbnail of S_KOM_2107922_Chapter4.pdf] Text
    S_KOM_2107922_Chapter4.pdf
    Restricted to Staf Perpustakaan
    Download (1MB)
    [thumbnail of S_KOM_2107922_Chapter5.pdf] Text
    S_KOM_2107922_Chapter5.pdf
    Download (266kB)
    Official URL: https://repository.upi.edu/
    Item Type: Thesis (S1)
    Additional Information: https://scholar.google.com/citations?user=o4wm3aIAAAAJ&hl=en ID SINTA Dosen Pembimbing: Yudi Wibisono: 260167 Eddy Prasetyo Nugroho: 5990993
    Uncontrolled Keywords: Fine-tuning, Laporan Tahunan, Large Language Model, LoRA, Peringkasan Teks Otomatis. Annual Report, Automatic Text Summarization, Fine-tuning, Large Language Model, LoRA.
    Subjects: L Education > L Education (General)
    Q Science > QA Mathematics > QA75 Electronic computers. Computer science
    Divisions: Fakultas Pendidikan Matematika dan Ilmu Pengetahuan Alam > Program Studi Ilmu Komputer
    Depositing User: Muhammad Rizki
    Date Deposited: 04 Sep 2025 04:49
    Last Modified: 04 Sep 2025 04:49
    URI: http://repository.upi.edu/id/eprint/137406

    Actions (login required)

    View Item View Item