Machine Learning Techniques for Depression Analysis on Social Media- Case Study on Bengali Community

Abstract

Depression is a prevalent illness in today’s society. It changes and influences our entire method of thought and our emotional, cognitive, and everyday behavioral behaviors. It affected over 264 million people, and the proportion increases every day. Mainly when it lasts for a prolonged time, it becomes a severe issue or health topic. It leads the trustworthy person to also malfunction, and that person commits suicide in his final position. There are several causes for depression, though social networking like Facebook, Twitter, and other networking plays a critical role in getting us more depressed. Most people in Asia use Facebook, Twitter, and various chat applications, and there they express their emotions. That is why our research initiative picks social media. Some work has been done on depression but depression detection on the Bengali community is done very rarely. So it has become a strong demand for today. The social media has intialted a study based on depression, tweets, and numerous chat app responses, and gathered Bengali data and projected depression posts and commentaries. Diverse approaches of machine learning have been used to evaluate these data and forecast depression and for algorithm purpose Support vector machine, Random Forest, Decision Tree, K-Nearest Neighbors, Naive Bayes (Multinomial Naive Bayes), Logistic Regression has been used. The desired results can be obtained by adding those algorithms. Moreover, different algorithms send us different results as trends were common, but ultimately the precision was the same for all algorithms applied to our dataset.