An Analysis of Machine Learning Algorithms for Chronic Kidney Disease Prediction

Chronic Kidney Disease (CKD) is a critical global health concern, especially in low- and middle-income nations, where compromised kidney function hinders effective blood filtration. Early detection is imperative as 10% of the world's population suffers from CKD, leading to increased fatalities due to limited access to affordable treatment. Over the past two decades, CKD- related deaths surged by 82.3 %, highlighting its significant contribution to global mortality, as per the 2010 world-wide Burden of Disease Study. The challenge is compounded by CKD's often asymptomatic early stages, emphasizing the need for routine testing as a vital diagnostic tool. Early identification is crucial in preventing End-Stage Renal Disease (ESRD), saving lives, and facilitating efficient treatment. This study focuses on developing a CKD predictive model using machine learning methods, analyzing a dataset with 9993 rows and 25 columns containing key kidney health information. Various methods, including Random Forest, Logistic Regression, Decision Tree, Support Vector Machine (SVM), k-Nearest Neighbours (KNN), and Naive Bayes, are compared for optimal CKD prediction. Addressing missing data ensures result accuracy. The ultimate goal is to provide a reliable, cost-effective model for early CKD identification, aiding medical professionals and patients by expediting diagnosis and enabling timely intervention.