abstract
- The surge of social media usage and its importance on political campaigns across the globe has been imminent. As new electoral processes begin, new social media tactics appear each time. This behavior does not exclude Latin American countries and candidates. In response, recent research has analyzed how these platforms may aid in the search for new and scalable methodologies to model voting preference from social media like traditional pollsters do. However, most of the literature does not consider multiple networks and delimits its scope to the analysis of sentiment classification in either Facebook or X (formerly Twitter). In this way, the current research aims to study how mostly ignored social media apps like Instagram and YouTube can change the outcome of predictions online. We consider the case of the 2024 Mexican Presidential election and how social media behavior relates to what the polls say about the candidates. We conclude that Instagram and YouTube can turn failed predictors into success by considering the importance that features from these platforms provide to machine learning models. © 2024 held by the owner/author(s).