A Machine Learning Framework for Multi-Stage Prediction of Chronic Kidney Disease Based on a New Dataset from Iraq

Abstract

Chronic kidney disease (CKD), is described one of the deadliest and outspread diseases in current time. CKD is hence known as the silent killer because it triggers decline in kidney function without any presenting signs and symptoms from first to end stage. The disease is comprised of five stages according to the glomerular filtration rate (GFR). Multi class, machine learning (ML) models including Logistic Regression (LR), AdaBoost and Support Vector Machine (SVM) were used for prediction of CKD stages. The new data set that collected for 750 patients in Al-Ramadi Teaching Hospital involves 15 clinical features. The performance of the models was measured in terms of accuracy, precision, recall, specificity, F1 score and AUC. The performance measures showed that the LR model had the highest overall accuracy rate for all stages <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(72.89 \%)$</tex>, and was followed by AdaBoost <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(68.44 \%)$</tex> and the SVM (63.11%). The minimum execution time was 0.04 s for the LR model, demonstrating that it could potentially predict the CKD stage in real time.