Tri-Algo guardian ensemble approach for fake news detection in social media

Abstract

Abstract The rapid spread of misinformation in the digital age highlights the need for effective fake news classification systems. This research evaluates classification methods using multiple datasets, which present distinct challenges and features. Advanced preprocessing techniques enhance data quality, while various word embedding methods capture linguistic and contextual features. To address class imbalance, oversampling and class weighting strategies are applied. An ensemble approach integrates multiple models to improve classification accuracy. Results indicate that advanced embeddings achieve superior performance, while traditional methods remain computationally efficient. Ensemble models consistently outperform individual classifiers, demonstrating their robustness and scalability. These findings emphasize the potential of deep learning techniques for real-world fake news detection.

Keywords

• Tri-Algo guardian
• Passive aggressive classifier
• Stochastic gradient descent
• Logistic regression
• Fake news detection