Enhanced Prediction of Chronic Kidney Disease using XGBoost Machine Learning Model

Chronic kidney disease (CKD) might progress to end stage renal disease; moreover, cardiovascular dangers are dire. Machine learning used in for more speed and accurate diagnosis of CKD. The CKD prediction model proposed in this paper was developed using the XGBoost algorithm, which is quite effective in classification problems. Other clinical parameters such as blood urea, serum creatinine and white blood cell count are some of the 24 indices identified from among the 400 patient records in the dataset. Feature selection using SelectKBest was relevant, and hyperparameter tuning was done by RandomizedSearchCV Both quantitative and categorical data were preprocessed. Altogether, 75% of data used for training, while 25% of data used for testing. The XGBoost model had a better result with 96.88 % recall, 100% precision, and 98% accuracy. However, the proposed approach has disadvantages; namely, a small sample cross-section and possibly an imbalanced class. Further, the dataset will be increased, the methods of dealing with class imbalance will be applied using SMOTE algorithm, and the effectiveness of the proposed model will be tested in real clinical practice. This work also highlight how crucial it is to employ and enhance machine learning, especially XGBoost to detect early stage of CKD, proper treatment, low mortality rate, and increased survival rate among patients.