Comparative Analysis of Machine Learning, Deep Learning, and Large Language Models for Fake News Detection

Fake news presents a critical challenge to public trust and social stability in the digital era. This study evaluates the effectiveness of three computational approaches which are Machine Learning (ML), Deep Learning (DL), and Large Language Models (LLMs) for fake news detection. The objective is to determine the most effective approach by analyzing accuracy, computational efficiency, and scalability. Using a benchmark dataset comprising fake and authentic news articles, the models were trained and evaluated after applying comprehensive data preprocessing techniques to ensure effective text representation and feature extraction. The results indicate that ML models offered simplicity and high efficiency, while DL models, such as BERT, achieved competitive accuracy but demanded substantial computational resources. LLMs, though slightly less competitive than BERT, also exhibited significant resource requirements. These findings highlight critical trade-offs between accuracy, efficiency, and resource demands, providing valuable insights for selecting optimal models for real-world fake news detection systems.