Predictive models for early detection of chronic diseases in elderly populations: A machine learning perspective

Abstract

Many patients do not realize they have chronic kidney disease (CKD) until it is advanced. Early identification is crucial. CDK is a global health issue, especially for elderly people, requiring early and correct diagnosis to optimize treatment outcomes. This study proposes a VGG16 DL model for automated early detection of chronic kidney disease (CKD) using the CKD dataset. The model uses advanced data preparation techniques like feature selection, normalization, and missing value management to improve input quality. With a 98% F1-score, 98% accuracy, 97% precision, and 99% recall, the proposed VGG16 model exhibits remarkable performance. It is compared against baseline ML models, such as Random Tree, XG Boost, as well as Decision Tree, all of which perform worse. These findings demonstrate how well the VGG16 model captures intricate clinical patterns, offers a strong and trustworthy tool for CKD prediction, and facilitates better patient care by enabling prompt data-driven decision-making.