HPSO-Optimized Support Vector Machine Model for Fake News Detection in Online Social Networks

The exponential growth of social media platforms has enabled the rapid dissemination of information; however, it has also led to the proliferation of fake news, which can manipulate public opinion, cause panic, and disrupt social harmony. This study presents a machine learning-based framework for fake news detection in online social networks using an Optimized Support Vector Machine (O-SVM) classifier. The proposed model integrates Term Frequency–Inverse Document Frequency (TF–IDF) and Word2Vec embeddings for text representation, while a metaheuristic feature selection algorithm—Hybrid Particle Swarm Optimization (HPSO)—is employed to enhance the performance of SVM by selecting the most discriminative features. The model was trained and evaluated on the MediaEval 2019 dataset, comprising 12,000 news items labeled as real or fake. Experimental results demonstrate that the O-SVM model achieves a classification accuracy of 97.8%, outperforming baseline models such as Logistic Regression (91.2%), Random Forest (94.6%), and conventional SVM (95.1%). Precision, recall, and F1-score metrics also exhibit significant improvement due to feature space optimization. The robustness of the proposed system was validated through k-fold cross-validation and comparative analysis across multiple text embedding techniques. The study concludes that optimized machine learning models, particularly SVMs with adaptive feature selection, can effectively mitigate misinformation on digital platforms, offering a scalable and interpretable solution for real-time fake news detection.