MACHINE LEARNING ALGORITHM SELECTION FOR CHRONIC KIDNEY DISEASE DIAGNOSIS AND CLASSIFICATION

Abstract

In last decades, chronic kidney disease (CKD) becomes a global health problem that is steadily developing worldwide. It is a chronic illness highly related to increased morbidity and mortality, cardiovascular diseases, and high healthcare cost.  Earlier identification and classification of CKD is treated as a major factor in controlling the mortality rate. Data mining (DM) techniques are used for the extraction of hidden details from the clinical and laboratory patient data that is used to aid doctors in enhancing diagnostic accuracy. Recently, machine learning (ML) techniques are commonly employed for the prediction and classification of diseases in healthcare sector. With this motivation, this study examines the performance of different ML algorithms to diagnose CKD at the earlier stages. The proposed model involves data pre-processing in two stages such as missing value replacement and data transformation. Besides, a set of five ML based classification models are involved such as support vector machine (SVM), random forest (RF), logistic regression (LR), K-nearest neighbor (KNN), and decision tree (DT). For investigating the performance of the different ML models, a benchmark CKD dataset from UCI repository is employed and the results are examined under different aspects. Among the different classifiers, the RF model has accomplished superior results with the maximum precision of 0.99, recall of 0.99, and F-score of 0.99 with a minimal error rate of 0.012.