Depression Detection on Social Media Twitter Using Hierarchical Attention Network Method

Abstract

Mental illness, including depression, is not a mild condition that only some mentally weak people experience. Technology is developing so rapidly, especially communication technology through social media. Twitter is a very popular social media today. Users can easily quickly and simply communicate all the feelings they are experiencing through tweets, which allows us to find information about emotional feelings to the level of user depression. Auto-mated analysis of social media has the potential to provide a method for early detection. This study aims to predict early signs of depression using data from social media Twitter. The method used in this research is classification by analyzing social media sentiment using the Hierarchical Attention Network. Classification using the Hierarchical Attention Network method was chosen because the method showed outstanding results for classifying texts in previous studies. The classification model in this study that represents the best accuracy, 74%, was performed by applying the Hierarchical Attention Network.