The Early-stage Detection of Kidney Disease using Machine Learning

Abstract

Chronic kidney disease is one of the most widespread diseases on Earth, at any given moment, it is estimated that around 697.5 million people are diagnosed with CKD. There are 5 stages of CKD, each stage gets progressively worse, until stage 5, where dialysis or a kidney transplant is needed to maintain life. If detected in the early stages (stage 1-2), treatments can slow or even stop the progression of CKD. However, detecting early stages of CKD is very difficult because patients usually do not have symptoms until stage 4, by then, the kidney’s function is cut in half. In this paper, we demonstrated utilization of data efficient machine learning (ML) algorithms as Decision Tree, SVM, and Logistic Regression on publicly available data to predict CKD. Further interpretation of the algorithms revealed significance of certain attributes (e.g. blood pressure, age) and we were able to get 95 percent + accuracy using certain features about patients.