Abstract

The purpose of this study is to develop a more accurate, reliable, and scalable fake news detection system by using an ensemble machine learning approach that overcomes the limitations of traditional single-model classifiers, particularly low accuracy, poor generalization, and high false-positive and false-negative rates. The study adopts an experimental research design using a benchmark Kaggle fake news dataset consisting of 20,800 labeled news articles. Text data were preprocessed through tokenization, stop-word removal, Bag-of-Words representation, TF-IDF transformation, and N-gram feature extraction. Five machine learning classifiers Support Vector Machine, Logistic Regression, Random Forest, Decision Tree, and Gradient Boosting were trained individually and then combined into an ensemble model using a majority-voting mechanism. The system’s performance was evaluated using accuracy, precision, recall, and F1-score. The ensemble model significantly outperformed individual baseline models, achieving an accuracy of 97.86%. The results show that combining multiple classifiers reduces misclassification rates and enhances predictive reliability. The proposed system demonstrates strong robustness, adaptability, and effectiveness in detecting fake news across diverse textual inputs. The study relies on a single benchmark Kaggle dataset, which may not fully represent the complexity of real-world news across different platforms, languages, and evolving misinformation styles. The model focuses on textual content only and does not incorporate multimedia elements such as images, videos, or social context. This research contributes a novel ensemble-based framework that integrates multiple machine learning algorithms into a unified voting system for fake news detection. By combining diverse classifiers, the model significantly improves accuracy and robustness compared to traditional single-model approaches, offering a practical and scalable solution for real-world misinformation detection.