PERBANDINGAN NLP DAN SVM DALAM ANALISIS SENTIMEN KOMENTAR INSTAGRAM TERKAIT STIGMA MASYARAKAT TERHADAP BANJIR SUMATERA 2025

Abstract

This study aims to analyze public sentiment regarding the 2025 Sumatra flood based on Instagram comments using a Natural Language Processing (NLP) approach. The methods applied include IndoBERT and Support Vector Machine (SVM) with TF-IDF features. A total of 711 comments were collected through a crawling process and processed using preprocessing techniques. The results show that negative sentiment dominates at 44.7%, followed by positive (30.4%) and neutral (24.9%) sentiments. Model evaluation indicates that IndoBERT outperforms SVM with an accuracy of 74.8% compared to 66.4%. WordCloud visualization reveals dominant terms such as flood, government, forest, and palm oil, reflecting public concerns about environmental issues and government policies.