Ensemble Machine Learning for Anxiety, Depression, and Panic Attack Detection: A Stacking-Based Approach

Anxiety, sadness, panic attacks and other psychological issues greatly lead to negative conditions pertaining to mental health. Because of the overlapping symptoms, it gets highly difficult to distinguish them from one another. The dataset consists of numerical replies provided by 4,184 people diagnosed by a French university, and the study, addressed to people older than 18 years of age, sees this as its problem. The class imbalance issue was tackled using by the Synthetic Minority Over-sampling Technique (SMOTE) that guarantees all important classes are fairly represented. This was followed by classification performance improvement through a stacking ensemble learning framework that combines Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and AdaBoost classifiers. Results show that the best-performing psychological state predictors are the ensemble one, with highest accuracy against individual classifiers.