Fake News (Hoax) Detection on Social Media Using Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) Methods

Abstract

Information can be found and shared through social media effectively and quickly. One of the most widely used social media is Twitter. Any information shared on social media is not always true. With millions of social media users, the platform cannot be separated from disseminating information whose truth is uncertain. This has a negative impact on society because it can increase people's distrust of information circulating on social media. To overcome this problem, This research propose system that can detect hoax information on social media using deep learning. This research focuses on detecting hoaxes using the Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) methods using a dataset from Twitter of 25.325 data. To obtain optimum results, this study utilizes Feature Expansion in the form of GloVe (Global Vector) and Feature Extraction with TF-IDF (Term Frequency-Inverse Document Frequency). The uniqueness of this research lies in the combined application of TF-IDF feature extraction with GloVe feature expansion using CNN and RNN deep learning methods. The results of this study prove that the hoax detection system, by applying a combination of extraction feature with expansion features, can increase the accuracy value up to 95.09% in the CNN classification method, and in the RNN classification method, it has an accuracy of 95.12%.