This work presents a comparative study of machine learning, deep learning, and transformer-based models for automated fake news detection. With the rapid spread of misinformation on online platforms, developing reliable computational methods to identify misleading content has become increasingly important. In this study, multiple approaches are evaluated, including traditional machine learning models, deep neural network architectures such as Bidirectional Long Short-Term Memory (BiLSTM), and transformer-based models. The models are assessed on their ability to distinguish between real and fake news using textual features extracted from news articles. In addition to evaluating predictive performance, this work places strong emphasis on model interpretability. Post-hoc explainability techniques, including LIME (Local Interpretable Model-agnostic Explanations) and SHAP (SHapley Additive exPlanations), are employed to analyze model behavior and identify the most influential features driving predictions. This analysis provides insights into how different models make decisions and contributes to improving transparency in automated misinformation detection systems. The results highlight the strengths and limitations of different modeling approaches and demonstrate the importance of explainability for building trustworthy NLP-based fake news detection systems. The findings may inform future research on robust and interpretable methods for combating online misinformation.
Soumya Sinha (Mon,) studied this question.