Machine Learning Approaches For Detecting Fake News: Ensemble Stacking Outperforms Conventional Methods

Abstract

The fast propagation of fake news in social media seriously damages public trust and societal integrity. This paper presents NDetect, a stacking-based ensemble learning model designed for effective fake news detection. NDetect combines Logistic Regression, Decision Tree, Support Vector Machine, and Random Forest classifiers to provide additional robustness and predictive accuracy. The model was trained on the ISOT dataset, using TF-IDF features and standard preprocessing. In experimental results, NDetect performed superiorly against individual models, achieving 89% accuracy and 0.957 ROC-AUC score. In comparison with conventional methods, the ensemble approach exhibits very good generalization and classification performance. These results carry implications for the ensemble learning approach in developing a robust misinformation detection scheme on digital platforms.