Predicting digital engagement on Instagram
(A2020-63299)

AUTHORS

Lorenzo Vecchi, Pontifical Catholic University of Paraná, Curitiba Campus (PUCPR); Eliane Francisco-Maffezzolli, Pontifical Catholic University of Paraná, Curitiba Campus (PUCPR)

KEYWORDS

Engagement prediction; Artificial intelligence; Instagram

ABSTRACT

This work aimed to develop a method to predict the level of engagement of HEIs on Instagram with artificial intelligence. The final product, however, in addition to foresight, permeated ways to classify Instagram posts, analyze the relevance of using people for such feature extraction, and description of the patterns found by the program, due to the characteristic of being an interpretable machine-learning algorithm. The database used was collected from public images provided by one private Instagram HEI. The work has an exploratory character, despite having a step that aims to determine a possible causal relationship between the independent and dependent variables, whose final premise was the description and understanding of the patterns found. The results demonstrate the contribution that variables extracted by human perception may have, in addition to the different patterns learned for different engagement metrics.