Important Features Associated with Depression Prediction and Explainable AI

Depression is a debilitating disease that leaves individuals persistently feeling sad or hopeless for more than two weeks affecting more than 300 million people globally. We applied several machine learning models with model explain-ability to a publicly available depression dataset. Several experiments were performed to assess the use of feature selection methods and technique to address dataset imbalance on diagnostic accuracy. The top performing model was obtained by logistic regression with excellent performance metrics (91% accuracy, 93% sensitivity, 85% specificity, 93% precision, 93% F1-score and 0.78 Matthews correlation coefficient). Feature importance was also generated for the best model. Explainable artificial intelligence method using LIME was applied to help understand the reasoning behind the model's classification of depression leading to better understanding of physicians, thus demonstrating its use in clinical practice.