Analisis Sentimen Menggunakan IndoBERT-BiLSTM dan Pemodelan Topik Menggunakan BERTopic terhadap Program Makan Bergizi Gratis

Fiantono, Ellen Theodora (2025) Analisis Sentimen Menggunakan IndoBERT-BiLSTM dan Pemodelan Topik Menggunakan BERTopic terhadap Program Makan Bergizi Gratis. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002211019-Undergraduate_Thesis.pdf - Accepted Version Restricted to Registered users only Download (18MB) | Request a copy

Abstract

Program makan bergizi gratis (MBG) yang diinisiasi oleh Presiden Prabowo Subianto pada tanggal 6 Januari 2025 bertujuan untuk menyediakan makanan bergizi bagi anak-anak sekolah, balita, ibu hamil, dan ibu menyusui guna mengatasi masalah gizi buruk dan stunting di Indonesia. Untuk memahami reaksi masyarakat terhadap program makan bergizi gratis, penelitian bertujuan untuk menganalisis sentimen dan topik yang berkembang di media sosial, khususnya twitter atau X. Analisis sentimen dilakukan dengan menggunakan model IndoBERT-BiLSTM dan BERTopic untuk menggali perubahan opini masyarakat selama periode kampanye, setelah presiden terpilih, dan program berjalan. Hasil penelitian menunjukkan bahwa model IndoBERT-BiLSTM mampu menganalisis sentimen pada data test dan mendapatkan akurasi pengujian sebesar 92.75% pada dataset periode kampanye, 94.64% pada dataset setelah presiden terpilih, dan 96.44% pada dataset program berjalan. Metode pemodelan topik BERTopic ini didasarkan dengan IndoBERT Embedding. Dengan data tweet yang ada, model BERTopic mampu menghasilkan coherence score sebesar 0.6731. Nilai 0.6731 tergolong cukup baik dan menunjukkan bahwa model BERTopic mampu menangkap struktur tematik yang cukup jelas dalam data tweet yang digunakan. Hal ini juga menunjukkan keunggulan BERTopic dalam mengolah tweet berbahasa Indonesia.
=================================================================================================================================
The Free Nutritious Meal Program (MBG), initiated by President Prabowo Subianto on January 6, 2025, aims to provide nutritious food for school children, toddlers, pregnant women, and breastfeeding mothers in an effort to combat malnutrition and stunting in Indonesia. To understand public reactions to this program, this study analyzes sentiment and emerging topics on social media, particularly Twitter (X). Sentiment analysis was conducted using the IndoBERT-BiLSTM model, while topic modeling employed BERTopic with IndoBERT embeddings to explore shifts in public opinion across three periods: the campaign phase, post-election, and program implementation. The results show that the IndoBERT-BiLSTM model successfully classified sentiment with test accuracies of 92.75% during the campaign, 94.64% after the president was elected, and 96.44% during program implementation. Topic modeling using BERTopic achieved a coherence score of 0.6731, indicating a fairly strong semantic structure in the generated topics. This suggests that BERTopic effectively captured meaningful thematic patterns from Indonesian-language tweets, highlighting its advantage in processing public discourse on social media.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Analisis Sentimen, BERTopic, BiLSTM, Dynamic Topic Modelling, IndoBERT, Pemodelan Topik, Program Makan Bergizi Gratis, Twitter, BERTopic, BiLSTM, Dynamic Topic Modelling, Free Nutritious Meal Program, IndoBERT, Sentiment Analysis, Topic Modeling, Twitter
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Mathematics and Science > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Ellen Theodora Fiantono
Date Deposited:	31 Jul 2025 04:07
Last Modified:	31 Jul 2025 04:07
URI:	http://repository.its.ac.id/id/eprint/124044

Actions (login required)

View Item