Development of a computer system to screening patients with chronic kidney disease

This work aims to construct a computer system to aid in the early diagnosis of Chronic Kidney Disease (CKD) using noninvasive clinical data, exploring machine learning techniques. Data collection was performed at a referral center for treatment of chronic kidney disease. The database consists of 443 participants (instances), of whom 178 have no renal disease (control) and 265 have chronic kidney disease. The clinical data collected were: Gender, Age, Stature, Weight, Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP) and Diabetes. To classify chronic kidney disease, four classifier algorithms were tested: Random Forest (RF), Naive Bayes (NV), Support Vector Machine (SVM) and K-nearest neighbors (KNN). The classifier that obtained the best result was applied a graphical interface. Among the classifiers, the SVM showed more accurate results than the other classifiers with 93.18% of accuracy, the sensibility and specificity parameters were also higher than the other methods, 0.96 and 0.88, respectively. Thus, SVM was the classifier used to obtain the computer system that is available online for health professionals and the general population, presenting a low cost and easy execution alternative for screening patients with CKD.