Comparative evaluation of machine learning algorithms for Alzheimer’s disease detection

Alzheimer disease or Alzheimer (AD) is one of the highest causes of dementia and it affects millions of people all over the world and it is characterized by memory loss and cognitive decline brought about by abnormalities in the brain such as amyloid plaques and tau tangles. Recent developments in artificial intelligence (AI) provide potentials regarding early detection and diagnosis using non-invasive technology making its accuracy in diagnosis and patient outcomes high even though it has a challenge in its clinical practice. Authors of the current study compare different machine learning algorithms to be used to detect Alzheimer disease based on large numbers of features obtained based on the demographic, clinical, and behavioral characteristics of patients. The dataset retrieved by authors on Kaggle is processed using Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbors (KNN), Support Vector Classifier (SVC), Gradient Boosting and XGBoost. These models are evaluated based on the common metrics such as accuracy, precision, recall and ROC AUC and a closer comparison between the performance of each model before and after hyperparameter optimization. The efficacy of machine learning models is proven, as Random Forest, XGBoost, and Gradient Boosting have a more favorable accuracy rate of over 90% and hence the prospect of machine learning models in diagnosis of early Alzheimer. The other issue involved in the paper is the implication of the findings on clinical practice and implementation of machine learning technology in medicine as a form of early diagnosis.