A Comparative Analysis of Machine Learning Algorithms with Bagging and Boosting for Mental Health Assessment

This study explores machine intelligence algorithms for evaluating mental health conditions on distinct datasets. The primary goal is to identify effective algorithms for anticipating mental health concerns. The central objective is to determine the most productive machine learning algorithms in predicting mental health issues using provided datasets. Various algorithms – Logistic Regression (LR), k-Nearest Neighbor (kNN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), bagging, and boosting – were employed with diverse parameter settings. The OSMI (Ds1) and the District Mental Health Program dataset (Ds2) were used for experimentation. Among the algorithms tested, LR with an Optimal Threshold (LR-OT) achieved notable performance with accuracy, precision, recall, specificity, and F1-score ranging from 0.87 to 0.89 across both datasets. kNN with Optimized parameters (kNNO) achieved 90% and 91% accuracy rates for Ds1 and Ds2, respectively. SVM obtained average accuracies of 91% and 93% for Ds1 and Ds2, respectively, with specific variations showing superior outcomes. Notably, DT, RF, bagging, and boosting models exhibited R2 scores exceeding 0.70, with bagging and DT yielding the highest R2 score. The study's findings demonstrate that DT and RF models enhanced by bagging and boosting techniques outperform other algorithms in predicting mental health concerns using the provided datasets. This highlights the significance of employing these models for effective mental health assessment through machine learning.