Investigation of different machine learning algorithm performances in the classification of Alzheimer's disease

Alzheimer's disease is a progressive and irreversible neurodegenerative disorder that affects millions of individuals worldwide. Early diagnosis plays a critical role in slowing disease progression, optimizing treatment approaches, and maintaining quality of life. In this thesis study, EEG data obtained from individuals diagnosed with Alzheimer's disease and healthy control subjects were analyzed. Features extracted from EEG signals using the dispersion entropy method were utilized to compare the classification performance of various machine learning algorithms. Dispersion entropy and performance calculations were conducted using custom algorithms developed in the MATLAB R2018a environment, and the extracted features were applied to K-Nearest Neighbors , Decision Tree , and Random Forest models. The performance of the models was evaluated using metrics such as accuracy, precision, recall, F1 score, and ROC-AUC, and the generalization ability of each model was tested through 10-fold cross-validation. The findings revealed that the KNN algorithm was insufficient in distinguishing the patient class and exhibited random classification between classes (AUC: 0.50). In contrast, the Decision Tree and especially the Random Forest algorithm demonstrated more balanced classification performance; the Random Forest model achieved the highest results with 75% overall accuracy, 71% cross-validation accuracy, and an F1 score of 0.80. Furthermore, the dispersion entropy analysis indicated that entropy levels in the EEG signals of Alzheimer's patients were significantly reduced, particularly in the occipital, as well as parietal and temporal regions. These findings suggest that DE is capable of reflecting neurophysiological impairments and may serve as a potential EEG-based biomarker for the early diagnosis of Alzheimer's disease.