INTERNATIONAL JOURNAL FOR INNOVATIVE RESEARCH IN MULTIDISCIPLINARY FIELD

Abstract: The fast spread of information on social media has made detecting fake news a major challenge, as false content can sway public opinion and lower trust in online information sources. Traditional machine learning methods have been commonly used for classifying fake news; however, they often fail to fully understand the meaning of complex text. To overcome this issue, this paper introduces a hybrid model for fake news detection that combines both deep learning and machine learning approaches. The model uses Bidirectional Encoder Representations from Transformers (BERT) to get detailed semantic information from news articles, which is then used for classification with the XGBoost algorithm. The model is tested on a dataset of labeled fake news using metrics like accuracy, precision, recall, F1-score, and confusion matrix. The results show that the BERT + XGBoost model performs better in terms of accuracy and overall effectiveness compared to traditional machine learning models and standalone BERT classifiers. The study shows that combining contextual feature extraction with a strong gradient boosting classifier improves the effectiveness and reliability of fake news detection.