Détection de stress en utilisant l’apprentissage profond dans les réseaux sociaux

Abstract

Emotion analysis and emotional computing have attracted much interest in various research fields in recent decades, particularly with the emergence of problems concerning users' psychological health such as stress, anxiety and depression. To analyze these social media impacts, textual analyzes are particularly effective in identifying characteristics of human behavior and describing emotional state. In this project, advanced deep learning techniques to analyze social media data are used, in order to understand the emotional signals that present the stress indices expressed in texts. To detect stress in social networks, we analyzed textual data from Twitter and Reddit platforms. Using the LSTM model makes it possible to capture temporal and contextual dependencies in texts, and to accurately identify stressful emotions. Additionally, the LSTM model performance is compared with that of classical methods.