Hybrid Weighted-Voting Ensemble and XGBoost for Accurate Fake-News Classification

The aim of this study is to design and test an ensemble with weighted voting and XGBoost machine learning models on the balanced fake new dataset and compare their performance with classical models and previous work. The project applies an ensemble of supervised machine learning models and XGBoost on the balanced dataset of fake new, with text preprocessing and TF-IDF feature extraction. Both subjects achieve high levels of performance across all metrics, including high F1 scores through 5-fold cross validation and 80/20 Train-Test split. The developed models succeed on identification of fake news and can be applied to different datasets for further testing. The system demonstrates robustness and generalization but can be improved through covering more languages and preventing topic drift in the system.