DX: Depression Detection System Through X Using Hybrid Machine Learning

Abstract

Currently, there is an increasing prevalence of depression among Thai people, often expressed through social media. Unfortunately, many individuals suffering from depression are unaware of their condition. This article introduces a depression detection system through X called D2X, which utilizes sentiment analysis of X users tweets to predict their level of depression. The D2X processes various types of tweets, including text messages, emoticons, and images, using a hybrid machine learning approach that combines support vector machine and random forest techniques. The study showed that combining text with emoticons resulted in the highest performance. Additionally, the research revealed that the most crucial feature for predicting levels of depression is the text tweets. Emoticon and image tweets were also found to enhance the effectiveness of detecting depression. The D2X model, utilizing all types of tweet data, achieved the highest F-measure compared to other machine learning techniques. However, when using only the text messages from tweets, the D2X model showed marginally lower performance than DistilBERT but outperformed other deep learning techniques. The D2X also had the least model construction and usage time.