An approach to the sequential evaluation of emotional behaviors of depressive users on social networks in groups and individually

• Ano de defesa: 2021
• Autor(a) principal: Giuntini, Felipe Taliar
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Affective computing; Análise temporal; Computação afetiva; Depressão; Depression; Mineração de padrões sequenciais; Redes sociais; Sequencial pattern mining; Social networks; Temporal analysis
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/55/55134/tde-18082021-093843/

The constant growth in the use and sharing of data on social networks has provided opportunities to develop intelligent solutions for understanding different dimensions of human behavior online since users share social aspects, feelings, and opinions daily. In this way, several studies in Affective Computing have been conducted to recognize and predict emotional and indicative aspects of mental problems through the mining of complex data, such as texts, images, videos, and emoticons, available in social network posts. Depression is a common and growing health problem globally and is considered the third largest cause of incapacity for work, and the leading cause of emergency in health centers is characterized by the manifestation of a set of symptoms for at least two weeks. Symptoms can be compounded by profound sadness, guilt, loss of pleasure and mixed and atypical characteristics, which may be correlated to contexts and severely impact various social aspects. Although it is necessary to observe emotional characteristics over time, as it is known in the literature, studies have focused on classifying whether a given post is depressive and have not addressed the temporal recognition of mood manifestations and aspects of personality context. This Thesis aimed to answer \"how to recognize temporal patterns of behavior of depressive users in online social networks?\" In this way, an approach for the temporal recognition of behavioral patterns of depressed users on social networks is presented, composed of two methodologies that allow (i) the temporal evaluation of the behavioral patterns of user interactions in groups combining modeling and metrics of complex networks and recognition of emotions and feelings, and (i) sequential recognition of the patterns of behavior of individual depressive users, through the mining of frequent patterns of emotional and contextual characteristics. Information from posts and comments was used in both methodologies, composed of texts and emoticons present in the users timeline. Through complex network measures and frequent pattern recognition, the approach was evaluated, indicating to recognize strong patterns of interactional, emotional, and contextual behaviors online over time, which serve as indicative for human behavior specialists and are based on evidence in the literature.