Using text analytics on social media posts to identify cues or features of depressive behavior

Abstract

Depression is a common mental disorder and becomes the leading cause of mental disability worldwide. World health organizations (who) in 2021 reported that nearly 700,000 people die due to suicide yearly. Delays in diagnoses or treatments, inaccuracies, missed diagnoses and therapies for depression are common issues. The possibility of social media as a tool for early depression intervention services to predict depression is examined as people often posts about their daily life on social media platforms. This is considered as a help-seeking behaviour because people who suffers from depression may have different intention to get support through their social network connection. By analysing the language cues in social media posts, machine learning models based on text analytics may be developed to provide an individual with information into his or her mental health earlier than conventional approach. Therefore, the aim of this study is to analyse the text data related to depression extracted from the social media posts to identify the cues or features of depressive behaviour in order to build an algorithm that can effectively predict depression. This study used the cross-industry process for data mining (crisp-dm) methodology for developing the depression detection model. The identified cues from the text analytics are presented as wordcloud. The results show that depressive people tend to have excessive sleep and those who are given a heavy workload often linked to negative emotions felt, such as anger and fear. While, the positive wordcloud include soothing words such as "thank", "love", "fun", "happy" and "game". These cues provide inputs that may be useful to assess individuals with and without depression on social media and can be further explored to develop the depression detection model which will be helpful to physicians and psychiatrists in diagnosing mental diseases and analysing patient behaviour.