Document Type : Research Paper


1 University of Tehran

2 School of Industrial Engineering, College of Engineering, University of Tehran, Tehran, Iran



Feature selection is the process of picking the most effective feature among a considerable number of features in the dataset. However, choosing the best subset that gives a higher performance in classification is challenging. This study constructs and validates multiple meta-heuristic algorithms to optimize machine learning models in diagnosing Alzheimers. This study aims to classify cognitively normal, mild cognitive impairment, and Alzheimers by selecting the best features. The features include Freesurfer features extracted from MRI images and clinical data. We use well-known machine learning algorithms for classifying, and after that, we use multiple meta-heuristic methods for feature selection and optimizing the objective function of the classification. We consider the objective function a macro-average F1 score because of the imbalanced data. Our procedure not only reduces the irreverent features but also increases the classification performance. Results show that metheuristic algorithms can improve the performance of machine learning methods in diagnosing Alzheimers by 20%. We find that classification performance can be significantly enhanced by using appropriate meta-heuristic algorithms. Meta-heuristic algorithms can help find the best features for medical classification problems, especially Alzheimers.


Main Subjects