Predictive Modeling of Misinformation Diffusion Patterns on Facebook and X in the 2025 Philippine Midterm Elections

Abstract

Social media platforms such as Facebook and X play a central role in political information dissemination in the Philippines, particularly during election periods when the rapid spread of misinformation can influence voter perception and democratic processes. This study examines the structural diffusion patterns of verified election-related misinformation during the 2025 Philippine midterm elections and develops predictive models for early classification of amplification behavior. 65 misinformation cases, involving approximately 20,000 posts, were quantified using amplification metrics, including size, depth, width, reach, and speed. Unsupervised clustering methods-K-Means++, Ward's Linkage, and Average Linkage-were applied to standardized structural features, with validation metrics supporting a five-cluster solution representing distinct diffusion archetypes: low-amplification, layered constrained, deep high-intensity, broad shallow, and temporal resurfaced cascades. For predictive modeling, the generated clusters were used to train Logistic Regression, Random Forest, and Support Vector Machine classifiers following feature selection via the Boruta algorithm, which identified size, reach, depth, and width as key predictors. Stratified 5-fold cross-validation demonstrated strong internal consistency, and evaluation on an unseen test set yielded 92.86% accuracy and a macro-averaged F1-score of 0.8939 across models. Logistic Regression was selected as the final model due to its interpretability and comparable performance. The findings indicate that misinformation diffusion patterns exhibit structurally predictable characteristics, suggesting that early detection based on cascade geometry is feasible. This integrated framework, combining network analysis, clustering, and supervised learning, provides actionable insights for election regulators, fact-checking organizations, and social media platforms to support timely intervention during high-stakes democratic events.