Abstract
Chronic kidney disease is one of the critical illnesses that affects roughly 10% of the people in the world. Early and accurate prediction of such disease is required for proper treatment. The use of machine learning (ML) for medical diagnosis in healthcare has increased. The doctor can identify the disease early with the aid of ML algorithms and approaches. This study aims to develop a diagnosis approach to recognize chronic kidney disease and assist the experts for exploring preventive measures early using extreme gradient boosting (XGBoost) model. The XGBoost is used due to its ability in-build features to manipulate missing data and its regularization capability to handle unbalanced datasets. The approach is trained and evaluated on a public dataset consisted of 24 features for 400 patients taken from the University of California Irvine (UCI) repository. The mean and most frequent values are used respectively for replacing the missing numerical and categorical values. The experimental results using a 10-fold cross -validation and holdout test techniques with a number of evaluation metrics exposed that the XGBoost model of the proposed approach achieves a competitive high result compared with the recent work on the same dataset. It attained 99.9% of AUC mean for the 10 -fold cross-validation test and 99.6 of accuracy for 60% holdout test from the dataset to diagnosis the chronic kidney disease.