Enhanced XGBoost-Based Automatic Diagnosis System for Chronic Kidney Disease

Abstract

Chronic kidney disease is a very prevalent ailment in the world; National Kidney Foundation of South Africa estimated that about 15% of the population in South Africa experience kidney disease and about 20,000 yearly reported cases and several thousands die untimely due to this disease. Application of Artifical Intelligence (AI) techniques to our day-to-day lives is bring positive changes, from banking to health carem military, gaming, welfare and so on. Scholars have worked extensively on Chronic Kidney Diseases (CKD) and most of their works are on pure statistical models thereby creating a lot of gaps for Machine Learning (ML) based model to explore. In this paper, we will review existing techniques, and propose a better technique based on Extreme Gradient Boosting (XGBoost) model with a combination of three feature selection technique for a fast and accurate diagnosis of CKD with relevant symptoms. The CKD model developed in this paper has an accuracy of 0.976, which is better than the baseline models currently existing. Also, the sensitivity and specificity of the CKD model for 36 patients is 1.0 and 0.917 respectively. False diagnosis of CKD patients using this model is reduced greatly. The proposed model will reduce the cost of diagnosing CKD and it can be easily embedded in a CDSS.

Ähnliche Arbeiten