Analisis Sentimen Berbasis Aspek Menggunakan Hybrid IndoBERT-Conv1D-BiLSTM Pada Program Makan Bergizi Gratis

Dimas Aditia Anugerah Setiady
Sunneng Sandino Berutu (Universitas Kristen Immanuel Yogyakarta)
Yo'el Pieter Sumihar (Universitas Kristen Immanuel Yogyakarta)

Algoritme Jurnal Mahasiswa Teknik Informatika
Vol 7 No 1 (2026): October 2026 || Algoritme Jurnal Mahasiswa Teknik Informatika

This study analyzes public sentiment toward the Free Nutritious Meal Program (MBG) using an Aspect-Based Sentiment Analysis (ABSA) approach on X (Twitter) data. A hybrid IndoBERT-Conv1D-BiLSTM model is proposed to classify sentiment across six MBG aspects. A total of 16,227 tweets were collected and automatically labeled using an Indonesian RoBERTa model, resulting in 14,756 tweets for training. After Random Oversampling, the model achieved a validation accuracy of 91.82%, a test accuracy of 91.10%, and a macro F1-score of 0.91. Compared with IndoBERT-only, CNN-only, BiLSTM-only, and SVM with TF-IDF, the proposed model achieved the best performance. The findings indicate that the proposed model is effective for aspect-based sentiment analysis and supports more comprehensive policy evaluation.