Detection of Stress Level in Social Media users using Cnn-Fg Model

Abstract

Psychological stress has become a common condition in today's world owing to the busy life style and competitive environment. This has led to increase of suicidal rates in the recent years. Lately, there has been a tremendous increase in interactions in the social networking sites. As people are spending long hours in the virtual world it is easier to detect and analyze the stress levels of the social media users. In this paper, we have proposed a hybrid approach which is a combination of Factor Graph (FG) model and Convolutional Neural Network (CNN) to analyze the textual contents in social media users' tweets and posts to detect the level of stress of a user. The tweets of an individual user are gathered from Twitter platform which is preprocessed and passed to the cross autoencoder embedded CNN Model which outputs user level attributes. These are then input to the Factor Graph model that detects the stressed tweets. A mechanism has been proposed to inform the friends or relatives of the concerned stressed user if the detected stress level is above the given threshold.