A Hybrid Approach for Weak Learners Utilizing Ensemble Technique for Alzheimer’s Disease Prognosis

Abstract

Objectives: To develop a hybrid machine learning (ML) model that predicts Alzheimer's disease (AD) accurately. Methods : This study used the Open Access Series of Imaging Studies (OASIS) dataset to develop a hybrid ML model. Given this data, we utilized five algorithms i.e., Logistic Regression, Gaussian Naive Bayes, K Nearest Neighbor, Support Vector Machine, and Decision Tree. An ensemble technique was employed to construct an ML-based hybrid model with 343 observations, 40% of which were used for training and 60% for testing. Findings: Using the voting classifier technique, the hybrid Machine learning model obtained an accuracy of 89.28%. Following hyperparameter tuning, the model's accuracy was increased to 90.62%. The effectiveness of AD classification was assessed using Accuracy, Precision, Recall, and F1-score. Novelty: The results demonstrate that, even with a limited amount of training data, the Hybrid ML modelling approach can reliably predict Alzheimer's disease in real-world community settings.