Detection of Kidney Diseases: Importance of Feature Selection and Classifiers

Abstract

Chronic kidney disease (CKD) is a medical condition characterized by impaired kidney function, which leads to inadequate blood filtration. To reduce mortality rates, recent advancements in early diagnosis and treatment have been made. However, as diagnosis is time-consuming, an automated system is necessary. Researchers have been employing various machine learning approaches to analyze extensive and complex medical data, aiding clinicians in predicting CKD and enabling early intervention. Identifying the most crucial attributes for CKD diagnosis is this paper's primary objective. To address this gap, six nature-inspired algorithms and nine machine learning classifiers were compared to evaluate their combined effectiveness in detecting CKD. A benchmark CKD dataset from the UCI repository was utilized for this analysis. The proposed model outperforms other classifiers with a remarkable 99.5% accuracy rate; it also achieves a 58% reduction in feature dimensionality. By providing a reliable, cost-effective tool for early CKD detection, the authors aim to revolutionize patient care.