Classification of suicidal ideation severity from Twitter messages using machine learning

Abstract

Depression has become a major mental health problem in Thailand and can lead to suicidal ideation. As suicidal ideation may vary in intensity and lead to suicide attempts, early detection of suicidal ideation severity should be implemented. This research presents text classification models for the prediction of suicidal ideation severity. A dataset of Twitter messages in Thai was used to develop several classification models. A web application prototype was also developed to predict suicidal ideation severity and introduce self-therapy based on Cognitive Behavioral Therapy to its users for managing negative automatic thoughts. The application prototype received satisfactory feedback during the user experience assessment. The results of this research highlight the importance and need for socio-technical systems to help with early suicidal ideation detection and early therapy in the social environment where mental health support is inadequate.

Keywords

• Machine learning
• Suicidal ideation
• Cognitive behavioral therapy
• Socio-technical system
• Social media
• Mental health