Abstract
As campaigns ramped up for the Philippine presidential election on May 9, 2022, the voting public was eager to learn more about the candidates’ platforms and political positions. Social media can be an instrument in gauging interest in the candidates and determining the topics or issues that resonate with their followers.
This study aims to obtain insights on what topics or issues resonated with the popular candidates’ followers by analyzing and identifying patterns in their Twitter engagement. There are two types of machine learning models created: 1) five binary classifiers that identify themes with high and low levels of engagement per candidate, and 2) a multiclass classifier that predicts which candidate is most likely to post a certain tweet. Topics that positively and negatively influence the prediction are then made clearer using interpretability techniques.
The resulting topics revealed that each candidate has their unique set of keywords that distinguishes them from one another. Followers of the candidates are more likely to engage in two types of tweets: personal and appreciative, and expressing dissent against the administration. The benefit of these insights is twofold: voters can gain a better understanding of what truly matters to the candidates, and the presidentiables can understand what their supporters anticipate from them.
