Alzheimer's disease prediction model using demographics and categorical data

Abstract

Diagnosing Alzheimer's disease (AD) is usually difficult, especially when the disease is in its early stage. However, treatment is most likely to be effective at this stage; improving the diagnosis process. Several AD prediction models have been proposed in the past; however, these models endure a number of limitations such as small dataset, class imbalance, feature selection methods etc which place strong barriers towards the accurate prediction. In this paper, an AD prediction model has been proposed and validated using categorical dataset from National Alzheimer's Coordination Center (NACC). The different categories such as Demographics, Clinical Diagnosis, MMSE & Neuropsychological battery, is preprocessed for important features selection and class imbalance. A number of predominant classifiers namely, Naïve Bayes, J48, Decision Stump, LogitBoost, AdaBoost, and SDG-Text have been used to highlight the superiority of a classifier in predicting the potential AD patients. Experimental results revealed that Bayesian based classifiers improve AD detection accuracy up to 96.4% while using Clinical Diagnosis category.