FNMD: An Evaluation of Machine Learning and Deep Learning Techniques for Fake News Detection

Fake news proliferation on social media platforms has become alarming because it poses threats to various aspects of society. Fake news encompasses intentionally falsified information designed to mislead readers and manipulate public perception, resulting in low-quality and misleading content. This paper presents a machine learning-based approach for fake news detection using: 1) Sentiment knowledge to calculate a sentence sentiment score as one of the features for sentence-level classification. 2) Statistical and linguistic knowledge to extract required features. The proposed method combines several types of features into a unified feature set to design a more accurate classification system (“True”: the fake news; “False”: otherwise). Thus, to achieve better performance scores, we carried out a performance study of five well-known feature selection techniques and several most famous classifiers to select the most relevant set of features and find an efficient Machine Learning (ML) classifier, respectively. Alongside traditional machine learning models, we also applied different Deep Learning (DL)-based models. Different machine learning and deep learning-based methods are applied to three different datasets including Liar, GossipCop, and PolitiFact, and the results show the integration of machine learning-based classification method and feature selection technique can significantly improve the performance and make the method comparable to other existing methods. Furthermore, DL models, in particular LSTM, exhibited remarkable accuracy thresholds of up to 88% in capturing intricate patterns within textual data.