An Explainable Stacking Ensemble Model for Alzheimer’s Disease Diagnosis

Alzheimer’s Disease (AD) is an irreversible neurodegenerative disease that causes a global health concern and affects millions of people. Current diagnostic approaches, such as neuroimaging techniques, are highly invasive, expensive, and not easily accessible. Gene expression profiles from blood offer a less invasive and lower-cost alternative. This study proposed a stacking ensemble model that enhances the predictive performance of AD diagnosis. Feature selection techniques such as Random Forest (RF), Recursive Feature Elimination (RFE), and Elastic-Net regression were used to select the optimal features for AD. Random Search hyperparameter tuning was employed to fine-tune the parameters of the proposed model. The proposed model incorporated Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naıve Bayes (NB), and Decision Tree (DT) as base learners, with Gradient Boosting (GB) as the metalearner. The Shapley Additive Explanations (SHAP) technique was utilized to enhance the interpretability of the proposed model. The proposed model achieved an accuracy of 0.9333 (95% confidence interval [CI]: 0.8667 – 0.9867), a Precision of 0. 9378(95% CI: 0.8847 – 0.9871), a Recall of 0.9333(95% CI: 0.8667 – 0.9867), and an F1-score of 0.9340(95% CI: 0.8701 – 0.9867).