Detecting Mental Disorders through Social Media Content

Abstract

Mental illness affects millions of people around the world. The popularity of social media platforms and their rapid insertion into nearly all the facets of our lives have not ceased to increase. The abundance and availability of social media content in conjunction with Machine Learning can aid the development of a suicide and depression detector by uncovering specific behavioral cues of individuals from their online posts. The study consists of building an application that uses a deep neural network model trained on the collected dataset to help create a prediction model in real-time. This application acts as a monitoring tool that can help in reducing the effects of mental illness by early detection. In this article, we developed six deep learning models in which half of them were trained with word embedding. Results demonstrated that the CNN+LSTM with word embeddings achieved the best performance with an accuracy of 97.56% after 15 epochs, followed by the LSTM model with 97.48% accuracy.