Detecting Fake News using a Hybrid BERTXG-Boost Model

The sudden growth of online platforms has intensified the dissemination of misleading content, necessitating advanced detection mechanisms. This study introduces a hybrid classification framework that combines the linguistic depth of BERT with the robust decision-making capabilities of the eXtreme Gradient Boosting (XGBoost) algorithm. BERT is employed to extract nuanced contextual representations from news articles, which are subsequently utilized by XGBoost for final classification. Experiments on a balanced dataset of genuine and fabricated news articles reveal that the proposed approach surpasses conventional models in terms of accuracy, precision, recall, and F1-score. The findings highlight the model’s promise for real-world applications in identifying online misinformation.