Abstract :
The existence of the news circulating has a threat as a fake news and disturbing the public or often called a hoax. The application of machine learning can be used to determine or classify a news story as true or a hoax. Research for the classification of hoax articles has been carried out by Dina Maulina and Rofie Sagara using the Support Vector Machine (SVM) classification algorithm combined with word weighting with the Term Frequency – Inverse Document Frequency (TF-IDF) method. In addition, Aini Hanifa and others have also conducted a comparison of the Recurrent Neural Network (RNN) algorithm method between Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU). This study aims to compare the accuracy values between SVM and RNN algorithms to classify hoax news and determine the effect of a clickbait in classification. The dataset used is the result of scrapping from the website turnbackhoax.id with a total of 500 news data. The method used in this research is to convert the news dataset into a vector form with the Word2Vec method, which is then modeled using the SVM and RNN algorithms. The results of this study indicate that the RNN algorithm has a higher accuracy value when compared to the SVM algorithm for the classification of hoax news. The accuracy value obtained by the SVM algorithm is 88% for news in the non-clickbait category and 90% for news in the clickbait category, while the accuracy value obtained by the RNN algorithm is 92% for news in the non-clickbait category and also for news in the clickbait category. This study also shows that both algorithms have a higher accuracy value for news that has a clickbait category compared to those that do not or non-clickbait, which means that the clickbait category of news has an influence on news classification. Keywords : Hoax, Machine Learning, Classification, SVM Algorithm, RNN Algorithm, Algorithm Comparison.