Enhancing Veracity: Empirical Evaluation of Fake News Detection Techniques

Abstract

The relentless surge in the dissemination of fabricated news articles and misleading information poses a substantial threat to the veracity of information, trust in media sources, and the very foundations of democratic processes. In the contemporary landscape of information dissemination, the rampant propagation of false and misleading information has emerged as a paramount concern for both individuals and society at large. Consequently, the development of robust and efficient detection mechanisms has taken on a pivotal role in contemporary information ecosystems. The methodology employed encompasses three pivotal stages: initial data preprocessing, subsequent feature extraction involving the utilization of TF-IDF and Count Vectorization, and eventual classification that draws upon a diverse array of machine learning classifiers, in addition to deep learning models like Recurrent Neural Networks (RNN) and BERT networks. On a carefully selected dataset of instances of fake news, these models underwent rigorous testing and training. This research greatly aids in the continuous battle against disinformation by offering a methodical framework for identifying fake news. It also helps people and organisations make better decisions by shedding light on the relative merits of various deep learning and machine learning methods in this intricate and crucial context.