Optimized Novel Text Embedding Approach for Fake News Detection on Twitter X: Integrating Social Context, Temporal Dynamics, and Enhanced Interpretability

In the era of widespread misinformation, detecting fake news has become a crucial challenge, particularly on social media platforms. This paper introduces an optimized approach for Fake News Detection, combining BERT and GloVe embeddings with Principal Component Analysis (PCA) and attention mechanisms, enriched by social and temporal features for more effective text representation. Leveraging the CIC Truth Seeker Dataset 2023, we applied SHAP for feature selection and interpretability, ensuring transparency in the model's predictions. Our methodology achieved a remarkable accuracy of 99.9% using a Random Forest classifier, showcasing the efficacy of this optimized hybrid approach. The integration of interpretability techniques such as LIME and SHAP provides deeper insights into the model's decisions, making it a reliable tool for combating misinformation. This novel approach offers a robust and transparent solution to the growing threat of fake news, contributing significantly to the integrity of online information and public discourse on platforms like Twitter X.