Optimized Biomarker-Based Machine Learning Model for Accurate Early Detection of Chronic Kidney Disease

Chronic Kidney Disease (CKD) is a progressive and life-threatening disorder affecting millions worldwide. Early detection plays a critical role in preventing complications such as renal failure, cardiovascular disease, and increased mortality. Traditional diagnostic approaches rely primarily on serum creatinine and estimated Glomerular Filtration Rate (eGFR), which may not effectively detect early-stage CKD. This study proposes a robust machine learning framework integrating denoising and sparse feature selection approaches with biomarker analysis to enhance CKD detection accuracy. A Denoising Sparse Auto-Encoder (DSAE) is utilized to eliminate noise and extract meaningful representations from high-dimensional clinical data. Term Frequency–Inverse Document Frequency (TF-IDF) is adapted for identifying significant biomarkers. Sequential Selection Ratio (SSR) and Feature Normalization Component (FNC) techniques optimize feature selection and scaling. Finally, an Adaptive Backpropagation Neural Network (ABPNN) performs classification. Experimental results demonstrate superior accuracy, improved ROC performance, and enhanced diagnostic reliability. The proposed framework provides an efficient and scalable solution for early CKD detection and supports clinical decision-making.