Trustworthy Explainable Hybrid Framework for Reliable Fake News Detection

Abstract

The widespread use of social media and online news platforms has significantly increased the dissemination of misinformation, creating a growing need for reliable fake news detection systems. Although traditional machine learning techniques provide efficient classification, they often struggle to capture contextual meaning, while transformer-based models offer superior semantic understanding at the cost of higher computational complexity and reduced interpretability. To overcome these limitations, this study proposes a trustworthy and explainable hybrid framework that combines TF-IDF feature extraction, contextual BERT embeddings, Attention-BiLSTM, ensemble learning, and SHAP-based explainable artificial intelligence (XAI). By integrating statistical and contextual representations, the proposed framework enhances classification performance, robustness, and transparency. The Attention-BiLSTM module effectively captures contextual dependencies, while ensemble learning improves prediction stability and generalization. SHAP explanations provide insights into the features influencing classification decisions, thereby increasing model interpretability and user trust. Experimental results on the Fake and Real News Dataset demonstrate outstanding performance, achieving 99.82% accuracy, 99.76% precision, 99.71% recall, and 99.73% F1-score. Additional ablation, robustness, and cross-domain evaluations further confirm the effectiveness and adaptability of the proposed framework. The findings indicate that the proposed approach offers a reliable, interpretable, and practical solution for intelligent fake news detection in modern digital environments.