Advancing Fake News Detection using Hybrid Deep Learning with FastText and Explainable AI

The dissemination of fake news across digital platforms has become a critical challenge, affecting public opinion, social stability, and decision-making processes. Various machine learning and deep learning approaches have been proposed in the detection of fake news, but most suffer from limited contextual understanding and interpretability, hindering their applicability in real-world applications. This paper presents a hybrid deep learning framework, CNN-LSTM, for effective capturing of both local linguistic patterns and longrange contextual dependencies in news articles. In addition, FastText word embeddings are used to enhance semantic robustness so that the model is capable of handling misspellings and rare word variations popularly encountered in deceptive content. Explainable AI is further implemented using LIME to obtain transparent and human-understandable justifications for model predictions. The proposed approach is evaluated on the WELFake dataset using standard metrics for classification. Experimental results depicted the hybrid model achieved a superior performance of 94.8 %, thus outperforming traditional and separate models of deep learning with a balanced precision and recall. The integration of explainability further ensures trustworthiness, making the proposed system suitable for deployment in high-stakes misinformation detection scenarios.