Enhancing Fake News Detection via PSO-Optimized Ensemble Learning: A Comparative Study of SVM, NB, and RF

Given the rapid spread of fake news across digital platforms, there is a pressing need for a reliable and efficient detection method.Current ensemble learning models often lack optimal weight tuning, limiting their performance in fake news classification tasks.To address this gap, we propose a Particle Swarm Optimization (PSO)-optimized ensemble model that integrates Support Vector Machines (SVM), Naive Bayes, and Random Forest (RF) classifiers using a soft voting strategy.Text data is preprocessed and transformed into numerical features using TF-IDF vectorization.The dataset, derived from the ISOT Fake News corpus, is split into training (80%) and testing (20%) subsets.Each base classifier is individually trained and evaluated, followed by the construction and assessment of an unoptimized voting ensemble.Subsequently, PSO is employed to fine-tune the weights of the base classifiers within the voting ensemble, enhancing overall prediction performance.The optimized model achieves 98.32% accuracy and an F1-score of 98.33%, outperforming both the unoptimized ensemble and standalone classifiers, as well as surpassing several state-of-the-art methods.This approach not only improves detection accuracy but also offers a scalable, interpretable, and effective solution to the fake news problem.Performance is evaluated using standard metrics such as ROC curves and confusion matrices, providing a comprehensive assessment of the model's reliability.