Predictive analysis of chronic kidney disease based on machine learning

The purpose of this study is to explore the influence of factors on patients with chronic kidney disease (CKD) and to establish predictive models using machine learning methods. Data were collected from the Affiliated Hospital of Nanjing University of Chinese Medicine between January 2016 and December 2017, including 69 CKD patients and 155 healthy subjects. This study found that carotid intima-media thickness (cIMT) is the most important indicator among the top 9 important features of each model. In order to find the best model to diagnosis CKD, extreme gradient boosting (XGBoost), support vector machine (SVM) and logistic regression are established and XGBoost is the most suitable model for this study (accuracy, 0.93; specificity, 0.89; sensitivity, 0.94; F1 score, 0.91; AUC, 0.99).