Performance Investigation of Different Boosting Algorithms in Predicting Chronic Kidney Disease

This paper implies an investigative approach to study the performance of different boosting algorithms in predicting chronic kidney disease (CKD) more accurately. In recent years, CKD has reached a global prevalence with much severity which can lead to end-stage renal disease (ESRD) if not detected early. Various boosting machine learning algorithms have been proven to be an effective tool to detect CKD in its initial stages. The dataset of the University of California, Irvine (UCI) repository has been utilized to train and test the model classifier containing 25 attributes. However, four different data frames were constructed by four different strategies (mean, median, mode, and null dropping method) to facilitate the missing values in the dataset. Eventually, three boosting algorithms were studied and corresponding confusion matrices are portrayed. Hence, a broad comparative investigation was conducted in terms of accuracy, precision, sensitivity, F1 score, ROC-AUC of each algorithm. Maximum accuracy of 99.75% was observed in the case of AdaBoost and LightGBM algorithms while 99.5% accuracy was noticed for XGBoost.