<![CDATA[In the 2024 election, widely celebrated as a festival of democracy, the spread of hoax news through social media has become a significant concern. A total of 203 hoax issues were identified, resulting in the spread of 2,882 hoax contents across various websites. A survey by Katadata Insight Center (KIC) revealed that between 30% to 60% of Indonesians were exposed to such hoaxes, with only 21% to 36% of them able to accurately discern whether the news was factual or a hoax. This research aims to develop a model using the K-Nearest Neighbour (KNN) algorithm to evaluate the accuracy and reliability of news from various Indonesian news websites through machine learning-based classification techniques. By categorizing news into potentially factual and potentially hoax categories, this study seeks to provide insights into which news pages are more likely to present factual information. The method employed involves Term Frequency-Inverse Document Frequency (TF-IDF) word weighting combined with KNN modeling, using Euclidean distance for calculation. The KNN model achieved a training accuracy of 0.90 and a testing accuracy of 0.88. The findings indicate that news sites such as detik.com, okezone.com, liputan6.com, and cnnindonesia.com demonstrate higher accuracy in the news they present. The use of the KNN method successfully identified probabilities of factual and hoax potential across news websites, including kompas.com, liputan6.com, detiknews.com, antaranews.com, cnnindonesia.com, okezone.com, sindonews.com, kumparan.com, pikiranrakyat.com, and wartatransparansi.com]]>
Copyrights © 2025
