UCLA

Alzheimer’s disease is a type of neurodegenerative disease that is common among the elderly. Although there is no cure, early diagnosis allows for treatments that can manage and delay the symptoms. We will employ machine learning algorithms, such as logistic regression, random forest, and extreme gradient boosting, to predict Alzheimer’s disease in two experiments. In the first experiment, each model is applied to all 450 features. In the second experiment, each model is applied to 25 different feature sets, with one set corresponding to each task. Predictions are based on the DARWIN (Diagnosis AlzheimeR WIth haNdwriting) dataset, and model performance is measured using accuracy, ROC curves, and AUC. The results indicate that the random forest model applied to all 450 features is the best performing model in predicting Alzheimer’s disease, achieving a model accuracy of 91.43% and an AUC of 0.9441.