<![CDATA[The escalation of geopolitical conflict between Iran, the United States, and Israel in 2026 triggered fluctuations in global oil prices, subsequently driving up domestic fuel prices in Indonesia and provoking various public responses on social media. This study aims to analyze the dynamics of public sentiment toward this issue and examine its temporal relationship with fuel price movements. The research data comprises 3,000 YouTube comments collected during the January–June 2026 period, classified using the IndoBERT model (indobenchmark/indobert-base-p1) fine-tuned with a Stratified 5-Fold Cross Validation scheme, Class Weighting, and Focal Loss to address class imbalance. A total of 900 comments were manually labeled as training data, while the remaining 2,069 comments were predicted by the model, resulting in 2,967 fully labeled data points. Weekly sentiment dynamics (n=24) were analyzed using four trend-testing methods (Linear Regression, Mann-Kendall, Spearman Correlation, and Chi-Square/Cramer's V), while the sentiment–fuel price relationship was tested using three approaches (Level Linear Regression, Cross-Correlation Function, and Differenced Regression). Results show the model achieved an average Macro F1-score of 0.7476 (SD=0.0297) with stable performance across folds. Sentiment distribution was dominated by the Negative class (57.1%), followed by Neutral (23.2%) and Positive (19.7%). Trend testing indicated a gradually improving sentiment trend, confirmed by three of four methods, while the relationship analysis found strong lag-3 correlations for three of five fuel types that did not persist under differenced regression, suggesting a possible spurious correlation rather than a confirmed causal relationship.]]>
