• Pendaftaran
  • Webmail
  • Perpustakaan
  • Portal Akademik
  • Universitas
  • Indonesia
    • English
Universitas Gadjah Mada Departemen Ilmu Komputer dan Elektronika
Fakultas Matematika dan Ilmu Pengetahuan Alam
  • Home
  • Departemen
    • Manajemen
    • Staf Administrasi & Teknis
    • Fasilitas
    • Kontak & Lokasi
  • Program Studi
    • S1 Ilmu Komputer
    • S1 Elektronika dan Instrumentasi
    • Magister Ilmu Komputer
    • Program Magister Elektronika dan Instrumentasi
    • Magister Kecerdasan Artifisial
    • Program Doktor Ilmu Komputer
  • Riset
    • Kelompok Riset
      • Elektronika dan Instrumentasi
      • Sistem Komputer dan Jaringan
      • Algoritma dan Komputasi
      • Rekayasa Perangkat Lunak dan Data
      • Sistem Cerdas
      • Lab Riset​ Sistem Tertanam dan Robotika​
    • Roadmap Penelitian dan Pengabdian kepada Masyarakat
    • Laboratorium Layanan
      • Lab Komputer Dasar
      • Lab Elektronika Dasar
      • Lab Instrumentasi Dasar
  • Pengajar & Publikasi
    • Daftar Dosen
  • Akademik & Mahasiswa
    • Pendaftaran
    • Program Pra Doktoral
    • SOP Panduan MBKM DIKE
    • Laporan Evaluasi Semester Program Studi
    • Prosedur
    • Mahasiswa
      • HIMAKOM
      • HMEI
      • OMAH TI
    • Beasiswa
    • Dokumen Akademik
  • Beranda
  • Semua Berita
  • Berita Program Sarjana
  • Berita Prodi S1 Ilmu Komputer
  • Mahasiswa Ilmu Komputer DIKE UGM Kembangkan Model Hybrid Deep Learning untuk Analisis Sentimen Berbasis Aspek

Mahasiswa Ilmu Komputer DIKE UGM Kembangkan Model Hybrid Deep Learning untuk Analisis Sentimen Berbasis Aspek

  • Berita Prodi S1 Ilmu Komputer, Kegiatan
  • 2 Juli 2025, 08.33
  • Oleh: wimbono
  • 0

Yogyakarta, 1 Juli 2025 – Peter Johan Arkadhira Setiabudi, mahasiswa Program Studi Ilmu Komputer, Departemen Ilmu Komputer dan Elektronika (DIKE) FMIPA UGM, berhasil mengembangkan model hybrid deep learning untuk meningkatkan akurasi dalam analisis sentimen berbasis aspek (Aspect-Based Sentiment Analysis/ABSA). Inovasi ini akan dipresentasikan dalam International Conference on Advanced Machine Learning and Data Science (AMLDS) 2025 yang akan diselenggarakan pada 19-21 Juli 2025 di Tokyo, Jepang.

Penelitian ini mengusulkan kombinasi BERT (Bidirectional Encoder Representations from Transformers) dan BiLSTM (Bidirectional Long Short-Term Memory) sebagai model hybrid untuk mengatasi keterbatasan pendekatan ABSA konvensional. Model dikembangkan dan diuji menggunakan dataset terbaru FABSA (Feedback Aspect-Based Sentiment Analysis) yang memuat lebih dari 10.000 ulasan dari berbagai sektor. Dengan menggabungkan keunggulan BERT dalam memahami konteks kata dan kemampuan BiLSTM dalam mempertahankan urutan kata, model ini mampu meningkatkan performa dalam memahami opini pengguna secara lebih mendalam. Penelitian ini dibimbing oleh Faizah, S.Kom., M.Kom., dosen DIKE UGM.

Hasil evaluasi menunjukkan bahwa model BERT-BiLSTM yang dikembangkan mencapai skor rata-rata F1 sebesar 0,7583 untuk tugas deteksi kategori aspek (Aspect Category Detection) dan akurasi sebesar 0,9334 untuk klasifikasi sentimen kategori aspek (Aspect Category Sentiment Classification). Nilai ini menunjukkan peningkatan performa dibandingkan model BERT standar, yang hanya mencatatkan F1-score sebesar 0,6425 dan akurasi 0,9112. Hasil ini menunjukkan bahwa pendekatan hybrid dapat meningkatkan kualitas analisis sentimen.

Analisis sentimen berbasis konteks menjadi salah satu pendekatan yang semakin relevan dalam memahami opini publik secara lebih mendalam untuk informasi digital saat ini. Teknologi ini memiliki berbagai potensi pemanfaatan, mulai dari pemantauan media sosial, analisis ulasan pelanggan, hingga pengembangan sistem rekomendasi yang mempertimbangkan emosi pengguna. Dengan kemampuannya dalam mengolah data teks yang kompleks, pendekatan ini dapat mendukung pengambilan keputusan yang lebih tepat oleh industri, pembuat kebijakan, maupun lembaga riset yang ingin lebih responsif terhadap berbagai pendapat masyarakat.

Penelitian ini juga memberikan kontribusi terhadap pencapaian Tujuan Pembangunan Berkelanjutan (SDGs), khususnya SDG 9: Industri, Inovasi, dan Infrastruktur. Melalui pengembangan solusi teknologi berbasis kecerdasan buatan,dan memperkuat ekosistem digital Indonesia.  Selain itu, penelitian ini juga membuka peluang eksplorasi lanjutan terhadap strategi penyeimbangan kelas data dan fine-tuning model untuk domain tertentu, sehingga dapat terus meningkatkan performa model dalam menganalisis opini masyarakat yang semakin kompleks. Diharapkan penelitian ini dapat memberikan salah satu kontribusi dalam pengembangan pemrosesan bahasa alami (Natural Language Processing/NLP) di masa depan.

 

Author: Peter Johan Arkadhira Setiabudi, Faizah

Editor: Marina 

#SDGs9

Tags: SDGs SDGs 9: Industri Inovasi dan Infrastruktur

Programs

  • Undergraduate Programs in Computer Science
    • + Regular Program
    • + International Program
  • Undergraduate Programs in Electronics & Instrumentation
    • + Regular Program
    • + International Program
  • Postgraduates
    • + Magister in Computer Science
    • + Magister in Artificial Intelligence
    • + Magister in Electronics & Instrumentation
  • + Doctorate in Computer Science

Prospective Students, welcome!


Berita UGM

  • UGM Student Team Promotes Ecotourism and Digital Village Development in Temanggung
    Universitas Gadjah Mada (UGM), through its Community Service Program (KKN-PPM), continues to demonstrate its commitment to regional development. UGM has deployed a KKN-PPM student team, Waksudha Bansari, to Bansari District, Temanggung Regency.  This time, the UGM team is collaborating with the Temanggung Regency Government to accelerate the development of ecotourism and digital agriculture. The collaboration […]
  • UGM Student Interns at Japanese Poultry Farm, Gains Insights into Precision Farming and Automation
    Muh. Dino Prasetyo (Dino), a student from the Faculty of Animal Science at Universitas Gadjah Mada (Animal Science UGM), has been interning at a layer poultry farm owned by Crest Co. Ltd. in Japan since February 2025. Throughout his internship, he has gained valuable experiences, particularly in time discipline and precision in performing tasks. At […]
  • UGM Professor Highlights Potential of Plant Bacteria as Natural Alternative to Fertilizers and Pesticides
    Just like humans, plants can also suffer from diseases caused by bacterial infections. The origins of plant bacteriology date back to 1878, marked by the discovery of fire blight disease in pear plants by Thomas Jonathan Burrill. Losses in agricultural production caused by bacterial plant diseases vary depending on the type of disease. Although economic […]
  • UGM to Conduct Internal Safety Review Following KKN Student Tragedy in Southeast Maluku
    Two Universitas Gadjah Mada (UGM) students, Septian Eka Rahmadi and Bagus Adi Prayogo, who lost their lives in a capsized boat accident in the waters of Debut, Manyeuw District, Southeast Maluku Regency, have been transported to their respective hometowns in Sumbawa Besar, West Nusa Tenggara, and Bojonegoro, East Java, on Tuesday, Jul. 1, 2025. Before […]
  • Dr. Adrianto Dwi Nugroho Inaugurated as UGM Professor, Highlights Justice in Global Minimum Tax Policy
    The Global Minimum Tax (GMT) initiative proposed by the Organisation for Economic Co-operation and Development (OECD) aims to create fairness in the tax system, particularly in ensuring that multinational companies pay a fair share of their tax obligations.  However, in recent years, aside from becoming an international economic and political commodity, the GMT policy has […]
  • Dr. Oyas Wahyunggoro Inaugurated as UGM Professor in Heuristic Control
    Dr. Oyas Wahyunggoro, a lecturer at the Department of Electrical Engineering and Information Engineering, Faculty of Engineering, Universitas Gadjah Mada (FT UGM), was inaugurated as a professor in the field of heuristic control on Thursday (Jun. 12). In his inaugural address, titled “Heuristic Control and Its Applications and Development Potential in the Automotive Industry,” Professor […]
  • UGM and Russia’s Positive Technologies Forge Cybersecurity Partnership
    Universitas Gadjah Mada (UGM) has opened opportunities for collaboration with Positive Technologies in areas such as academic capacity building, professional training, joint research, and national digital resilience programs.  This development emerged from a visit by a delegation from the Moscow-based global cybersecurity research company on Wednesday (Jun. 11) at the UGM Central Office. As an […]
  • UGM Campus Mosque Slaughters 50 Sacrificial Animals for Eid al-Adha
    UGM Campus Mosque received more than 50 sacrificial animals this year, donated by individuals and institutional partners. The total consisted of 11 cows, 19 goats, and 36 sheep. “We also received one cow from the rector’s family,” said the Head of the UGM Campus Mosque Takmir Board, Dr. Muhammad Yusuf, on Tuesday (Jun. 10). Dr. […]
  • Three UGM Medical Graduates Earn Perfect 4.00 GPA at Commencement Ceremony
    Three graduates from the Medicine Program at the Faculty of Medicine, Public Health, and Nursing (FK-KMK UGM) achieved a perfect cumulative GPA of 4.00 at the undergraduate and applied undergraduate commencement for the third period of the 2024/2025 academic year held on Wednesday, May 27, at Grha Sabha Pramana.  The three medical graduates are Claire […]
  • Dr. Suharyadi Inaugurated as Professor, Highlights Potential of Hybrid Remote Sensing Image Interpretation
    Optical remote sensing imagery has become an invaluable tool in urban area management, providing essential spatial data for urban environmental planning and management through its varied spatial, temporal, and spectral resolutions.  To identify objects on Earth’s surface in remote sensing imagery, it is necessary to analyze groups of pixels that form diverse spectral and spatial […]
Universitas Gadjah Mada

Departemen Ilmu Komputer dan Elektronika

Fakultas Matematika dan Ilmu Pengetahuan Alam

Universitas Gadjah Mada

Gedung C, Lantai 4

Sekip Utara Bulaksumur Yogyakarta 55281

Introduction

Mahasiswa

©

KEBIJAKAN PRIVASI/PRIVACY POLICY