Chronic Kidney Disease Prediction using Deep Neural Network

Abstract

Chronic Kidney Disease (CKD) is a global health issue and symptoms are not always visible at the early stage. Deep learning techniques can be developed to determine the factors that potentially cause CKD at an early stage to enable patients to receive timely treatment. This paper attempts to forecast Chronic Kidney Disease (CKD) by analysing a set of attributes. A publicly available dataset with information collected in India was used for carrying out the research. Data was first preprocessed using different techniques to deal with missing values and outliers in the dataset. Next, classification between CKD and notCKD was performed using both Random Forest and Deep Neural network. The results of both methods were compared, and it was found that the proposed DNN model yielded a superior accuracy of 98.8% for the binary classification.