Determining The Severity of Chronic Kidney Disease Using Machine Learning Methodologies

Abstract

Chronic kidney disease in short CKD is a worldwide health problem with high death rate and high morbidness, it can trigger other diseases too. CKD has no clear symptoms in its beginning phase, since patients fails to notice it earlier. Diagnosing the CKD in advance allows the patients to get the treatment on time to prevent the development of the disease. The data set was retrieved from UCI (University of California Irvine) machine learning repository. The data set which was obtained has a large number of missing values. Those missing values were filled using mean imputation method. After filling the missing values, two machine learning methodology such as K-Nearest Neighbor (KNN) and Logistic Regression (LOG) has been compared. On analyzing the performance of the two model KNN gives the better accuracy than LOG. Hence, the KNN model is proposed to diagnoses the presence and absence of CKD. The model is trained with 26 attributes and 400 samples. After diagnosing the CKD, the model predicts the severity of the disease by using eGFR (Estimated Glomerular filtration rate) value. This attribute helps the model to predict the stages of CKD which helps the patients to receive treatment in the early stage.