<![CDATA[This study aims to detect hoax news in Indonesian-language media by comparing two popular text classification methods: Naïve Bayes and Support Vector Machine (SVM). Unlike most prior studies that focus on English-language datasets, this research addresses a significant gap by analyzing hoax detection in the Indonesian context. The growing spread of misinformation online has made it increasingly difficult for the public to distinguish between factual and false information, often leading to anxiety, confusion, and social unrest. To tackle this issue, a dataset of 2,010 news headlines comprising 1,005 hoax and 1,005 factual titles was collected through web scraping from verified news portals and fact-checking websites. After undergoing text preprocessing and feature engineering using TF-IDF and N-Gram models, the data was classified using Naïve Bayes and SVM. Performance was evaluated in terms of accuracy, precision, recall, and computation time. The SVM model achieved 93% accuracy, 94% precision, and 93% recall, whereas the Naïve Bayes model yielded 93% across all three metrics. Notably, Naïve Bayes required only 5.2 seconds for classification, significantly faster than SVM's 15.7 seconds, highlighting a trade-off between speed and precision. A web application was developed using Streamlit to make the models publicly accessible, enabling users to test news headlines directly. This practical tool can assist journalists, fact-checkers, and policymakers in verifying information more efficiently. The findings confirm that both models are effective, with distinct advantages depending on the context of use.]]>
Copyrights © 2025
