<![CDATA[Abstract — This study aims to develop an Indonesian sentiment analysis system to examine public responses to the elimination of the Indonesian National Team in Round 4 of the 2026 World Cup Qualifiers. The method employed is a quantitative approach utilizing computational experiments based on an IndoBERTweet-SVM hybrid ensemble. The research stages comprise data collection, text preprocessing, feature extraction, model training, performance evaluation, and sentiment analysis on the target data. The tested models include a full SVM, SVM without pragmatic features, SVM without character features, IndoBERTweet transfer learning, and hybrid stacking. The evaluation results reveal that the full SVM yields the best performance with an accuracy of 0.7595 and a macro-F1 score of 0.7573. McNemar's test comparing the hybrid stacking to the best baseline model indicates a statistically significant difference with a p-value of 2.87e-08. The application of the model to 1,000 target data points demonstrates a relatively balanced sentiment distribution: 297 negative tweets (29.7%), 353 neutral tweets (35.3%), and 350 positive tweets (35.0%). These findings suggest that the response to the national team's failure is not entirely dominated by negative sentiments, but also includes evaluative, informative, and supportive elements. The study confirms that the SVM approach based on TF-IDF, character n-grams, and pragmatic features remains highly effective for the sentiment analysis of short Indonesian texts on social media. Furthermore, the developed system has demonstrated end-to-end analytical capabilities and can be further expanded for real-time data processing and broader digital sports studies. Key word — Sentiment Analysis; Indonesian National Team; Ensemble Hybrid; 2026 World Cup Qualifiers. ]]>
Copyrights © 2026
