Early Prediction of Chronic Kidney Disease Using Machine Learning Algorithms with Feature Selection Techniques

Chronic Kidney Disease (CKD) poses significant health risks, particularly for elderly and middle-aged individuals, leading to gradual kidney damage and reduced renal function. CKD’s impact on morbidity and mortality rates underscores the urgent need for early diagnosis. This study proposes a machine learning-based prediction system leveraging key physiological variables to accurately predict CKD stages. By employing feature selection techniques and various machine learning models, including Naive Bayes, K Nearest Neighbor, Support Vector Machine, Decision Tree, Random Forest, and Xgboost, CKD prediction was achieved with high accuracy. Notably, Random Forest Classifier and Xgboost attained 100% accuracy using 25 physiological attributes, while Chi-Square Test and Principal Component Analysis yielded a commendable 98.48% accuracy with only six attributes. Dominant features such as blood glucose random, serum creatinine, and hypertension were identified, facilitating efficient CKD prediction with minimal clinical data.