Perbandingan Labelling Menggunakan USRTerhadap Labelling Manual Untuk AnalisisSentimen Media Sosial Pada Kemacetan Lalu Lintas Di Kota Palembang

Abstract

This study aims to analyze public sentiment toward traffic conditions in Palembang City using comments from Instagram. The data were collected through web scraping from traffic information accounts, then cleaned and labelled using two approaches: keyword-based automatic labelling and manual labelling. The results show that most comments express negative sentiments, reflecting public dissatisfaction with traffic conditions. The consistency between automatic and manual labelling reached 62.02%, indicating that data balancing labelling is fairly effective but still needs contextual adjustments. In conclusion, social media–based sentiment analysis can serve as an alternative source of information for understanding public perceptions of urban traffic conditions