Análise de Traços de Personalidade em Mídias Sociais por meio de Redes Complexas

Abstract

The growth of social media and the volume of data generated by its users have sparked the interest of data science researchers in understanding behavioral patterns and psychological traits manifested in these environments. Much of the existing research focuses on using traditional classification models to predict user temperaments. However, these approaches have a significant limitation: they do not adequately capture the relationships and interactions between users, nor the dependencies between the extracted attributes. The use of complex networks seeks to overcome this limitation. This work proposes to analyze and correlate the temperaments/personality traits of social media users based on data obtained from these platforms, using complex networks. Unlike traditional classification approaches, the proposal aims to model textual and behavioral data extracted from the X platform (formerly Twitter) into network structures, enabling the detection of communities and the analysis of clusters of users with similar characteristics. Two databases containing posts from users labeled with personality types were used, undergoing stages of textual preprocessing, extraction of psychological attributes via LIWC, discretization, normalization, and network construction based on similarity measures. Analysis of the purest clusters revealed distinct patterns among the dimensions, especially Extraversion/Introversion (E/I) and Intuition/Sensation (N/S), suggesting that certain traits share linguistic and social characteristics observable in networks. The results support the hypothesis that complex networks are capable of identifying structural patterns associated with personality that complement traditional statistical analyses, contributing to new avenues of research.