AB-BiL: A Deep Learning Model to Analyze Depression Detection in Imbalanced Data

The usage of online resources (websites, social media, blogs, etc.) to express personal opinions is increasing daily. Annually, there are 300 + million people depressed worldwide. The objective of depression detection is to maximize the availability and accuracy of intervention. In existing works, depression detection has been done based on various machine and deep learning models. Previous depression detection utilizing social media-related data through deep learning techniques may help to bring changes in life trajectories and save lives. The major challenge of these models is the accuracy and lack of robustness and effectiveness in the models. The performance of the models is not satisfied because of the real-world imbalance present in the data distributions. To achieve this challenge, we are introducing a new model called AB-BiL. In our proposed method, we use deep learning models to increase the accuracy rate to detect depression in persons. The model involves a two-step process: (1) applying the AdaBoost model, which can be used to maintain the data imbalance. (2) A Bi-LSTM neural network can be used to increase the accuracy rate. Experimental results on the real-world datasets show that the proposed approach significantly maximizes the classification and prediction performance compared to the significant state-of-the-art techniques. The work represents the combined AdaBoost and Bi-LSTM network models to assist various mental health patients and practitioners.