Adarsh Maurya ¹, Arpit Pathak ² and Bhanu Priya ², ¹ Banaras Hindu University, India, ² Indian Institute of Information Technology, India

Chronic Kidney Disease (CKD) is a progressive medical condition that requires early detection to prevent severe health complications. In this work, we present a machine learning framework for CKD prediction that integrates feature selection and classification models. The UCI CKD dataset was utilized, and preprocessing steps included handling missing values, encoding categorical variables, and normalization. To reduce dimensionality and select the most relevant features, Particle Swarm Optimization (PSO) was applied, which significantly improved model efficiency. Multiple machine learning algorithms, including Random Forest (RF), Gradient Boosting (GB), Support Vector Machine (SVM), Logistic Regression (LR), and Artificial Neural Networks (ANN), were evaluated using 5-fold, 10-fold, and 15-fold cross-validation. The experimental results demonstrate that ensemble methods achieved the best performance after feature selection. Random Forest and Gradient Boosting consistently outperformed other models, achieving accuracies above 99% with near-perfect precision, recall, and F1-scores, while maintaining balanced classification between CKD and non-CKD cases. These results highlight the effectiveness of feature selection in improving diagnostic accuracy and confirm the superiority of ensemble learning for CKD prediction. The proposed approach provides a reliable and efficient tool that can assist healthcare professionals in early CKD detection and decision-making, ultimately contributing to better patient outcomes.

Keywords