An Explainable Ensemble Model for Depression Detection from Social Platform Data

Depression is a mental health condition marked by persistent feelings of sadness and a diminished capacity to enjoy or participate in usual activities. The detection of depression from social network data has attracted considerable interest due to the increasing prevalence of mental health issues and the widespread use of on-line platforms. Artificial Intelligence (AI) methods, leveraging Natural Language Processing (NLP) plays a crucial role in pinpointing potential risks. In this study, a deep integrated hybrid model consisting of Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) has been proposed for identifying depression from content shared on social platforms. The model has been trained and tested using a publicly available benchmark dataset consisting of posts from various social media platforms. Experimental evaluation and comparison with existing work and Machine Learning (ML) models has shown that the proposed model outperformed them all, achieving a remarkable accuracy of 97% on a large dataset, showcasing the model’s robustness and efficacy. Furthermore, to enhance interpretability and transparency of our model, we have applied Explainable AI (XAI), incorporating, LIME for local interpretability of the decision-making processes of our model.