A hybrid approach to fake news detection using FastText and explainable AI

The fast dissemination of untrue information through the Internet has weakened public faith in digital content alongside its general credibility. Modern systems for fighting fake news should analyse complete contextual signals because they detect misinformation that appears in multimedia content better than text-based analysis does. The authors present a detection system built around FastText and Convolutional Neural Networks (CNN) and the XLNet framework to increase accuracy level. XLNet utilizes transformer architecture to discover distant text correlations together with FastText which enhances word representation capabilities and CNN which flawlessly detects essential features from the text. Our hybrid system reaches superior classification performance when it combines these three consolidated components. The model possesses optimal performance because we adjusted its hyperparameters to stop overfitting thus making it work across multiple datasets. The evaluation process shows that the Fake Newsnet and LIAR detection model produces a 97% success rate in its identification tasks. We have implemented SHAP as an Explainable AI tool which explains how the model makes decisions through clear and easy-to-understand insights. The accessible display of the system allows users to trust the platform while understanding what factors lead to conclusions. This detection method establishes a strong decisive tool to identify fake news while maintaining complete system transparency which benefits falsehood identification efforts.