Uma infraestrutura computacional para a identificação de estudantes universitários com possível perfil depressivo usando dados de sensores móveis

Abstract

Depression, one of the most prevalent mental disorders, significantly impacts the lives of millions worldwide. This mental disorder not only causes severe health problems but also negatively interferes with the lives of those who suffer from it, many of whom do not seek treatment. In the university population, the prevalence of depression is even higher than in the general population. This work proposes a computing infrastructure that uses digital phenotyping data collected by mobile and wearable sensors, such as smartphones and smartwatches, to identify Brazilian university students with a possible depressive profile (PPD). Adopting the Design Science Research (DSR) methodology, the study develops a computational solution whose effectiveness is validated through a descriptive exploratory analysis of the collected data. This approach seeks not only to validate the feasibility of the proposed infrastructure but also to facilitate the identification of behavioral patterns associated with the PPD. The approach includes the use of the "Human in the Loop" method for removing data outliers and employs the t-SNE (t-distributed Stochastic Neighbor Embedding) technique for reducing high-dimensional data and identifying patterns in interpreting the results. This work aligns with the principles of Human-Computer Interaction (HCI) intending to contribute to the areas of mental health and computing.