Análises de publicações da rede social Instagram em contraste com o temperamento humano

Abstract

With the growth in the use of social networks, the possibilities for studies on social relationships and interactions have grown significantly. Understanding how users express their feelings and manifest their temperaments on social networks can be a step for anticipating psychological disorders. Instagram has billions of users and is the most used social networks today. However, this network is still little explored as a source of study for the human temperament. This work aims to analyze the relationships between the temperament of users and their data collected on the social network Instagram. For the analysis of the textual data of the users' posts, two sentiment classification strategies are used. The first one is an ensemble of classifiers and the second one is a layered classifier. Both strategies were evaluated in three databases and achieved accuracy results above 77%. The ensemble classifier obtained better results in the three bases with accuracy above 80%. For the analysis of the relationship between temperament and Instagram data, each user is represented by temperament (target attribute), number and average of likes, number of publications and proportions of use of emojis in publications, positive publications, negative publications , publications during the week, publications on weekends and followers. In this analysis, statistical tests, graphs and decision tree were used. The results indicate that users with a depressive mood post more positive-feeling captions than hyperthymic, irritable, and worried users. The behavior of publications during the week and on weekends of irritable and worried temperaments are similar, as well as the behavior of cyclothymic and depressive users. And finally, anxious users use more emojis in Instagram captions than depressed and angry users.