Machine Learning-Based Detection and Comparative Analysis of Fake News in Turkish and English

False information that spreads rapidly on digital platforms has the power to shape public perception and can lead to significant global consequences. The detection and prevention of fake news have become crucial to ensuring information reliability. In this study, we aim to detect fake news with high accuracy using machine learning models. We trained our models on multiple datasets, including both Turkish and English news sources, and evaluated various machine learning algorithms, including XGBoost, Random Forest, Naïve Bayes, Logistic Regression, and Support Vector Machines (SVM). Our models were assessed based on precision, recall, F1-score, and accuracy, and their performances were compared with each other as well as with findings from the literature. The results demonstrate the effectiveness of machine learning-based approaches for fake news detection and highlight the impact of language differences on model performance. We obtained our best results with SVM, achieving 94.1% accuracy for English news, and Random Forest, achieving 93.9% accuracy for Turkish news.