Enhancing Alzheimer’s disease diagnosis through metaheuristic feature selection and advanced classification techniques

Abstract

A diverse array of diagnostic and detection methods has been developed as a result of the advent of Alzheimer’s disease (AD) as a significant global health issue. This study employs bio-inspired algorithms, such as the parrot optimization algorithm (POA), grey wolf optimizer (GWO), and differential evolution (DE), to identify the most effective feature selection techniques for AD diagnosis. The predictive accuracy of these algorithms was improved by the simple keywords: Alzheimer’s disease optimization classification machine learning metaheuristic mentation of the Alzheimer’s disease Dataset. This was achieved by integrating a personalized fitness function and optimizing parameter settings with decision tree classifiers. To evaluate the algorithms’ effectiveness in machine learning models with population sizes of 30 and 60, precision, recall, accuracy, and F1-score were evaluated at 5, 15, and 30 iterations. The gradient boosting and XGBoost classifiers consistently obtained the highest results, while DE, GWO, and parrot optimization (PO) achieved maximal accuracy rates of 0.94, 0.93, and 0.94, respectively. These findings underscore the efficacy of integrating metaheuristic algorithms with robust classifiers to enhance the predictive accuracy of AD diagnosis. Furthermore, they illustrate that artificial intelligence (AI) algorithms that are operated by biological processes can accurately forecast AD, with the success rates and stability of the proposed methods serving as metrics for evaluating their efficacy.