ABSTRACT

This research addresses the critical problem of pervasive disinformation campaigns in Philippine elections, which significantly undermine democratic processes and public trust, especially given the nation’s high social media reliance and vulnerability to sophisticated manipulation. The escalating threat posed by generative AI and deepfakes further exacerbates this challenge, creating an urgent need for adaptive countermeasures. This study’s objective is to analyze historical and contemporary disinformation strategies and develop machine learning models capable of predicting these tactics. A mixed-methods approach was employed, combining quantitative analysis of social media data from the 2016 and 2022 election cycles with qualitative insights. Five machine learning algorithms – Logistic Regression, Random Forest, Support Vector Machine (SVM), Decision Tree, and Gradient Boosting—were trained and evaluated using metrics like Accuracy, Precision, Recall, F1-Score, and ROC AUC. Results indicate that SVM achieved the highest accuracy (0.7706) and F1-Score (0.8704), while the Decision Tree model demonstrated the best ROC AUC (0.5816), highlighting varied strengths across models in handling imbalanced datasets. These findings underscore the potential of machine learning to provide vital tools for early detection of disinformation, offering crucial insights for safeguarding electoral integrity in digitally connected democracies like the Philippines.

Keywords: Disinformation, Machine Learning, Philippine Elections, Social Media, Electoral Integrity

Author/s:

Huerte, Felix L. Jr.
flhuerte@nu-laguna.edu.ph
National University – Laguna

Claudel, Jeruz E.
National University – Laguna