A Two-Stage Text-to-Emotion Depressive Disorder Screening Assistance based on Contents from Online Community

Major Depressive Disorder (MDD) is one of the most significant medical problems. The total number of people living with depression in the world is more than 300 million. Nowadays, people use social communities to communicate with each other and express their mindset and emotion that are the sign of depression sealed under their feelings and could be the cause of suicide. This research proposes a two-stage predictive model for major depressive disorder risk screening assistance by using emotion values based on textual data from social community. The contents from social community are identified six emotional dimensions; angry, bored, excited, fear, happy, and sad. These emotions are then categorized into emotion-based clusters using DBSCAN algorithm and examined by medical and psychological experts with Patient Health Questionnaire-2 (PHQ-2). The labeled clusters is used to train and optimize classification model for each screening question with several algorithms; Decision Tree (DT), Random Forest (RF), and Gradient Boosted Tree (GBT). The experimental result shows that the GBT is the best model for Q1 and the RF is the best model for Q2 with 98.32% and 99.98% accuracy respectively. This research will be beneficial in the further study in identifying depression from text-based emotion.