Diagnosis of Chronic Kidney Diseases Using Machine Learning

Abstract

Chronic Kidney Disease has been affecting more than 10% of the world's population, and millions of people die each year due to the inaccessibility of affordable treatment. Chronic kidney disease is more common in older people, women, racial minorities, and people with diabetes and hypertension. According to data estimates, kidney disease affected nearly 843.6 million people worldwide in 2017. Early detection of CKD is important for preserving millions of lives. Many researchers have achieved substantial progress in this field by using various machine learning (ML) techniques to diagnose CKD at an early stage. However, more lightweight and cost-effective ML models are still mostly required. We provide a systematic and thorough strategy for dealing with difficulties, as well as a performance-optimized light-weighted medical data model. In this article, we used Random Value Imputation to replace missing values using the chi2 (chi-squared) technique to select features. Decision Tree Classifier, K-Nearest Neighbor, and Random Forest Classifier are the classification algorithms used in this study. Finally, the Random Forest Classifier can identify CKD with a 98% accuracy rate and no data leakage.