A Comparative Analysis of Machine Learning and Deep Learning Approaches to Enhanced Fake News Detection

Abstract

Fake news influences major aspects of society. To mitigate the dangers of misinformation, manual fact-checking is frequently implemented. Original fake news propagators target innocent individuals to disseminate misleading information. The government and society must first determine the pattern of false news dissemination to address this sequence of events. However, since manually fact-checking the substantial volume of newly generated data is insufficient, machine learning algorithms can be used to detect fake news on different social media platforms. This study applies and evaluates four machine learning algorithms and three deep learning models in detecting fake news. The area under the receiver operating characteristic curve (AUC), precision, F1-score, recall, and accuracy were used to determine the best algorithm for classifying fake news. Logistic Regression (LR) and Long Short-Term Memory (LSTM) both achieved the highest accuracy of 92.97%. This work elucidates the capacity of machine learning and deep learning models to detect fake news.