<![CDATA[The rapid growth of digital media in Indonesia has accelerated the spread of hoax news, which poses serious threats to public trust and social stability. Based on data from the Ministry of Communication and Informatics of the Republic of Indonesia, a total of 12,547 hoax contents were identified from 2018 to 2023, while 1,923 new hoax contents emerged in 2024 alone. This research aims to design and build a web-based hoax news detection system by implementing the Multinomial Naive Bayes algorithm combined with Natural Language Processing (NLP) techniques. The system processes text through five NLP stages: sentence splitting, case folding, tokenizing, stopword removal, and stemming using PySastrawi. Feature weighting is performed using TF-IDF (Term Frequency–Inverse Document Frequency), and classification is executed using the Multinomial Naive Bayes algorithm enhanced with Laplace Smoothing and Log Posterior Probability. The output is converted using the Softmax function to produce probability percentages for each sentence. A manual calculation simulation using 10 training sentences and one test narrative containing four sentences was conducted to verify the algorithm. The system successfully classified the test narrative as hoax with a probability of 62.00% hoax and 38.00% non-hoax, consistent with the actual label. The system was evaluated using a Confusion Matrix on a 50-sentence test dataset, achieving Accuracy of 90.00%, Precision of 91.67%, Recall of 88.00%, and F1-Score of 89.80%. The resulting system provides a practical tool for the public to verify the credibility of news information in the digital era.]]>
