Analysis and Implementation of a Hybrid Case-Based Reasoning and K-Nearest Neighbor Approach for Chronic Kidney Disease Prediction

Abstract

Chronic Kidney Disease (CKD) is a progressive deterioration of kidney function that frequently goes undetected in its early stages, posing a growing clinical concern — particularly among productive-age individuals whose diagnosis is often delayed until irreversible damage has occurred. Early and accurate prediction remains a pressing challenge, especially given the rising CKD incidence in this demographic linked to hypertension, diabetes, and shifting lifestyle patterns. This study developed a hybrid method combining Case-Based Reasoning (CBR) with weighted similarity and K-Nearest Neighbor (KNN) to improve prediction accuracy while preserving model interpretability. The dataset was obtained from the UCI Machine Learning Repository and filtered for productive-age individuals aged 15–64 years, yielding 288 instances after preprocessing. Attribute weighting was performed using Information Gain to reflect the varying diagnostic relevance of each variable, and inter-case similarity was measured through a weighted similarity approach. Classification was then carried out using KNN across multiple K values. At K = 2, the proposed method achieved an accuracy of 98.26%, with precision, recall, and F1-score each recorded at 0.983 — results that suggest the hybrid CBR-KNN approach is well-suited for deployment as a clinical decision support system for early CKD detection.