Depression Analysis from Social Media Data in Bangla Language: An Ensemble Approach

Abstract

Depression is a mental illness that has been harming individuals in their daily lives. With the advancement of technology, people rely on social media as means of communication. However, even though social media can significantly impact changing lives, the information from this platform is still considered vague and often disregarded. Moreover, with the hashtags and being on-trend, it is challenging to find depressive posts and help those in need. With the advancement of intelligence technology such as natural language processing and other machine learning algorithms, it has become easier to recognize patterns and ensure an effective digitized solution for depression analysis. There have been numerous studies about depression detection and analysis; however, most of them had not achieved a desirable outcome. Our paper intends to propose a model with a new approach for analyzing depression from Bangla social media posts. In our model, we have proposed a modified feature selection method along with different ensemble learning techniques. We have evaluated the performances of these techniques and acquired that the eXtreme Gradient Boost (XGB) Classifier with a 92.80% accuracy is the most suited for our model.