Predicting the Risk Factor for Developing Chronic Kidney Disease Using a 3-Stage Prediction Model

Abstract

Chronic Kidney Disease (CKD) is considered one of the major high-risk chronic diseases on humans’ health that causes death in its late stages. Moreover, treating CKD patients cost huge amounts to be paid based on their stage. In fact, it becomes significant not only to detect the disease in its early stages, but also, to have a way to early assess and predict the possibility for individuals to get affected in the future. In this research, a 3-Stage predictor is introduced to help predicting the risk factor for developing CKD during the healthcare screening based on a questionnaire and some laboratory tests. Also, it aims to reduce and eliminate the unjustified tests’ costs unless the tests are needed for the assessment by categorizing parameters into stages. A comparison between 12 classifiers led to choosing the 3 classifiers used in designing the 3-Stage model, based on the best accuracy and prediction speed. The 3-Stage model is designed using Bagged, Boosted and Medium Trees classifiers. The model was assessed on the dataset collected from the Centers of Disease Control and Prevention (CDC) in the United States. The trained 3-Stage model resulted in 99.97% accuracy by predicting around 3K cases in comparison with a 1-Stage model.