Hybrid Machine Learning Model for Detecting Depression on its Early Stage

The proposed study presents a simplified and yet innovative idea of addressing the early detection problem for depression from social media post. Review of literature suggest different variants of machine learning model towards detection of depression; however, there are open-end issues. Hence, the current work introduced a hybridized machine learning model using Gradient Boosting, Random Forest, and Support Vector Machine. The model also contributes to a novel manifold feature extraction approach to enrich the final feature vector quality. The prime contribution of this study is its improved generalization, adaptability, and scalability. Tested with massive benchmarked dataset of social media, the proposed model exhibited 97% accuracy and 25% faster response time in contrast to standalone version of existing machine learning model.