A Machine Learning Approach for the Early Detection of Dementia

Abstract

Longer life spans in today's society have contributed to the growth of degenerative disease prevalence, especially dementia. Dementia causes a deterioration in thought process and a decline in cognitive function, specifically thinking, reasoning, and remembering. While dementia cannot be completely prevented, its early detection can delay the onset of the disease. With the help of a machine learning algorithm, relevant attributes to detect the disease in its early stages can be refined and successful predictions can be made. To conduct this analysis, the Alzheimer Features and Exploratory Data Analysis for Predicting Dementia datasets were utilized. The following machine learning models were applied to the dataset: Naïve Bayes, Decision Trees, K-Nearest Neighbors, and Fully Connected Neural Networks. After evaluation of accuracy scores, confusion matrices for both Naïve Bayes and Decision Trees were determined to provide the best results among the models. While further investigation with a larger dataset is necessary, such models suggest that machine learning algorithms are a promising tool to detect and mitigate the growth of dementia in older populations.