An Artificial Neural Network Framework to Predict Patients with High Likelihood of Chronic Kidney Disease

Abstract

Chronic kidney disease (CKD) is an important health and healthcare system problem. The ability to predict which patients will develop CKD is a difficult task due to the complex nonlinear relationships among related factors. Using artificial neural networks (ANN), applied to a population 17 through 90 years of age, we achieved 97% accuracy in classification, based on standard laboratory test and patient data. The technique was also helpful in determining which features of the data are most predictive; 75% of the features were sufficient to reach this high level of accuracy.