hoax-news-indonesia

Amid the growth of online news and the risk of misinformation, this study developed a model to identify public sentiment, categorize news, and detect hoaxes. Unlike previous studies that used single models, the study concludes that the used of RoBERTa and Random Forest significantly improves accuracy in both sentiment analysis and hoax detection in Indonesian news. Using a Kaggle dataset, the models were trained and tested. For sentiment analysis, the RoBERTa model achieved 94% accuracy, with an inference time of 2 minutes and 40 seconds. For the hoax detection task, the Random Forest model outperformed Logistic Regression, achieving 96.92% accuracy, compared to 93.97% for Logistic Regression. The study concludes that the used of RoBERTa and Ensemble Learning significantly improves accuracy in both sentiment analysis and hoax detection. This can carry out the development of more informative, accurate and balanced digital media, particularly for Indonesian news.