Sentiment Analysis of US-China Tariffs using IndoBERT and Economic Impact on Indonesia

Abstract

The US-China trade war has influenced public perception due to its potential economic impact on developing countries like Indonesia. This study analyses Indonesian sentiment towards the tariff policies and their correlation with economic indicators. The dataset consisted of 38,739 social media comments collected through web scraping. The data were processed through data cleaning, case folding, stopword removal, normalization, and stemming. Each comment was labeled as positive, negative, and neutral. The dataset was split into 80% training and 20% testing sets, followed by an oversampling process to balance the class distribution. The data is fine-tuned using the IndoBERT model with the Python programming language. The model achieved its highest performance with an accuracy of 93.03%, precision of 93.42%, recall of 93.03%, and F1-score of 92.94%. Spearman correlation revealed a weak to moderate positive and significant correlation (ρ = 0.434, p-value < 0.05) between public sentiment and global soybean prices. These findings underscore the effectiveness of combining a deep learning model like IndoBERT with statistical analysis to link digital discourse to tangible economic indicators, highlighting the method's potential as a data-driven tool for policy evaluation.