Depression Detection from Social Media Data using CNN and Linguistic Metadata Features

Depression is a major cause of disability and a major contributor to the worldwide burden of mental diseases. Research shows that women are more prone than men to be depressed. More than 700,000 people kill themselves each year. Suicide ranks as the fourth leading cause of death for individuals aged 15 to 29 years. In this research, we employ several Machine Learning algorithms derived from messages and postings on social media networks to facilitate the early detection of melancholy. WORD2VEC, GLOVE, the Machine Learning algorithm LOGISTIC REGRESSION, and the neural network algorithm CONVOLUTIONAL NEURAL NETWORK (CNN) are trained and compared to classification-based user-level linguistic input.