Supervised Learning Approach for Detection of Sinhala Depressive Posts based on Twitter

Abstract

Depression is a common mental disorder and a treatable mental illness under a low cost if diagnosed at its early stage. But many of the affected people do not diagnose the disease at its early stage especially in Sri Lankan community due to the lack of awareness about the depression, having negative perception about the mental health services and social stigma. In serious stage, suicide could be a result of untreated depression. Therefore, early detection of depression plays an important role in psychology domain. Nowadays people are more open on social media platforms and tend to share personal information such as emotions, feelings, and problems. Also, they even expecting online help and guidance to overcome from their problems. As a result of that, social media has gained the attention of many researchers recently to detect depression using social media data. Even though this is a well-researched area using high resource languages like English. According to the author’s knowledge this is the first study which has been done to identify depressive contents in Sinhala contents. From this research, system was developed naming "DepDetect" which is based on Twitter platform. Prediction model was built and tested with five supervised machine learning algorithms (SVM, Multinomial Naïve Bayes, Random Forest, Decision Tree, KNN) and KNN was used to develop ‘DepDetect’ which observed the highest accuracy score with 70%.