Applying Data Mining Approaches for Chronic Kidney Disease Diagnosis

Abstract

Kidney disease is one of the most common problems today that many people in the world deal with it. Therefore, in this study, our main objective is to use several computational-based algorithms to classify and diagnose Chronic Kidney Disease. The applied data in our study were publicly available data on chronic kidney disease. Eight classifiers were used to classify chronic kidney disease into two groups (patient or not). We used the Windows 10 operating system and RapidMiner Studio 9.8 version. The confusion matrix provides us the TP, FP, FN, and TN values; some performance measures were calculated to evaluate the used techniques. Evaluation of data mining algorithms revealed that Random Forest (with 100 trees), Deep Learning network (with five hidden layers), and Neural Network (with 0.02 training rate and 100 cycles) reached the highest accuracy rates with 99.09%, 98.04%, and 96.52% respectively. However, it is notable that Random Forest, Support Vector Machine, and Deep Learning network achieved 1 for AUC. Data mining on health-related issues can be considered one of the most useful data analysis tools. These classification methods are beneficial for specialists in the medical diagnosis process, and by using these techniques, hidden patterns are extracted from the raw data.