Detecting Depression From Social Media Using a Deep Learning Based Approach

Abstract

Unhappiness, feeling low, losing interest and excitement in daily life activities and it continues for a long time, causes depression. Depression can affect all types of age like young, adult and also child. Some signs and symptoms of depression are badly mood swings, loss of interest in daily life activities, less movement and speed, always feeling guilty and worthless, too much sleeping or Insomnia etc. Most of the people hesitate to talk about mental health. Nowadays depression is a common problem in our daily life. According to WHO, more than 264 million people are suffering from depression. The second major reason of death is suicide in the whole world and the age range is 15-29 years old and every year nearly 800000 people committed suicide due to depression. Researchers applied too many approaches to detect depression. But it is hard to understand someone’s emotions from social media. In this paper, a Bangla dataset collected from Facebook to detect depression. LSTM, CNN, Combined CNN-LSTM are applied to detect depression from text. Later, performance comparisons are shown of these three architectures. Hope that it will help psychologists and the other researchers in their work based on depression. It will help to prevent harmful behaviors which occur due to depression.