SED-MDD: A Novel Approach for Predicting and Assessing the Severity of Major Depressive Disorder Using Stacked Ensemble Learning

다국어 초록 (Multilingual Abstract)

The treatment of Major Depressive Disorder (MDD), a debilitating mental health condition affecting many individuals, necessitates precise diagnostic and prognostic approaches. With the rapid advancements in machine learning, innovative tools for mental health diagnosis and prognosis are being developed. The application of ensemble methods in MDD severity assessment has garnered recognition for enhancing predictive accuracy and reliability. This study introduces a novel approach, the Stacked Ensemble for Depression Management (SED-MDD), aimed at improving predictions of MDD severity using stacked ensemble machine learning techniques. At the beginning of SED-MDD processing data undergoes preprocessing through the combination of demographic information with clinical findings and EEG measurements. The method selects vital EEG indicators by using cluster-based feature selection to detect particular brain electrical activity patterns characteristic of depression. Cleaned and selected dataset goes through the prediction process when it is fed into multiple machine learning algorithms consisting of CatBoost and XGBoost. The combined machine learning approach accumulates favorable prediction results by achieving 96% accuracy in detecting MDD which exceeds standard predictive models. The model's ability to forecast data improves through the combination of health variables with EEG readings which enables personal care strategies during rehabilitation processes. The ROC analysis from the study proves that ensemble methods with feature selection techniques successfully spot severe MDD cases while emphasizing their critical role in improving patient outcome treatments.