Prediction of depression in social network posts using machine learning algorithms

Abstract

Depression is a serious mental disorder and its extreme or worst condition can lead to suicidal action. The number people who suffer from depression is drastically increasing day by day, particularly in teenagers who express it explicitly or keep it invisible which means the depressive feeling is hidden in deep down of their mind. Some of them manages to acknowledge it and some of them even do not know that they are in a depressed mindset. However, this feeling can be emitted in social media pool if the candidate has a habit of posting every event and situation on social networks. Depression silently kills may teenagers and their friends are unknown about it. Since many people maintain social network as an open diary and share everything related to their state of mind, the network users can have the possibility to know the partial scene of the user’s situation. If there is a system that can measure the level of depression from users’ continuous posts for a certain period of time and give an alert or pop-up notification to friends and family (followers), then we can save many young lives from this tragedy, Therefore, the objective of this work is to build a model by utilizing users continuous posts for a certain period of time. For this, we have investigated machine learning algorithms such as Naïve Bayes, Random Forest, Linear Regression, Support Vector Machine to select a best one with highest accuracy and Support Vector Machine performed better with highest classification performance for the prediction.