Analisis Sentimen Masyarakat terhadap Layanan Publik pada Media Sosial X Menggunakan IndoBERT dan Support Vector Machine

Abstract

Social media platform X has become one of the primary channels for the public to express opinions regarding public services, making it a valuable source for evaluating government service quality. This study aims to compare the performance of Support Vector Machine (SVM) and IndoBERT in classifying public sentiment toward public services using the Figshare Twitter Dataset. The dataset was first filtered using public service-related keywords, followed by preprocessing steps including case folding, text cleaning, stopword removal, and tokenization. The SVM model employed the Term Frequency–Inverse Document Frequency (TF-IDF) feature extraction method, while the IndoBERT model was fine-tuned using a Transformer-based architecture. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. The experimental results show that IndoBERT outperformed SVM, achieving an accuracy of 91.89%, precision of 92.51%, recall of 92.56%, and F1-score of 92.54%. In comparison, SVM achieved an accuracy of 81.20%, precision of 78.02%, recall of 78.74%, and F1-score of 78.88%. These findings indicate that IndoBERT is more effective in capturing the contextual semantics of Indonesian-language social media texts than the TF-IDF-based SVM approach. The proposed approach can support the development of intelligent sentiment analysis systems for monitoring and evaluating public services through social media.

Keywords—Sentiment Analysis, Public Services, Social Media X, IndoBERT, Support Vector Machine, Natural