Early Depression Detection from Social Network Using Deep Learning Techniques

Abstract

Depression is a psychological disorder that affects over three hundred million humans worldwide. A person who is depressed suffers from anxiety in day-to-day life, which affects that person in the relationship with their family and friends, leading to different diseases and in the worst-case death by suicide. With the growth of the social network, most of the people share their emotion, their feelings, their thoughts in social media. If their depression can be detected early by analyzing their post, then by taking necessary steps, a person can be saved from depression-related diseases or in the best case he can be saved from committing suicide. In this research work, a hybrid model has been proposed that can detect depression by analyzing user's textual posts. Deep learning algorithms were trained using the training data and then performance has been evaluated on the test data of the dataset of reddit which was published for the pilot piece of work, Early Detection of Depression in CLEF eRisk 2017. In particular, Bidirectional Long Short Term Memory (BiLSTM) with different word embedding techniques and metadata features were proposed which gave good results.