SENTIMEN ANALISIS PERSEPSI PUBLIK DI MEDIA SOSIAL TERHADAP DAMPAK PERSAINGAN DAGANG CHINA- AMERIKA MENGGUNAKAN BERT

This study examines public perceptions of China-US trade competition through BERT-based sentiment analysis of comments on YouTube during the period 2023– 2025. The main objectives include mapping public sentiment patterns, evaluating the performance of the BERT model in opinion classification, identifying factors that influence sentiment, analyzing the dynamics of opinion change, and providing recommendations for policymakers. The analysis results show a predominance of negative sentiment in response to the conflict's impact on global economic stability. The BERT model demonstrates high performance in detecting negative comments (recall 0.97; F1-score 0.83), but it still has limitations in classifying positive and neutral sentiment due to imbalanced data distribution and linguistic expression similarities. These findings indicate the need to improve preprocessing strategies and data balancing to enhance fair sentiment representation. Methodologically, BERT has proven capable of understanding context and emotional nuances in text, making it a relevant tool for large-scale public opinion analysis. The results of this study provide important contributions to policymakers and economic actors in understanding the social impact of trade conflicts and formulating more appropriate strategic responses.
Keywords: BERT, sentiment analysis, trade competition, China-US, social media, YouTube comments, deep learning, public perception, data balancing, global economic stability