Early Detection of Chronic Kidney Disease (CKD) using Optimized Machine Learning Models

Abstract

Chronic kidney disease is sometimes abbreviated to "CKD." The term"CKD" generally refers to this ailment. The kidneys are affected by this disorder, whichis also known as chronic renal disease. The immense progress made in the area of machine learning and artificial intelligence is what has ignited the interest that has beenproduced as a result of these developments. Thus, any doctor with access to thedialysis report has the capacity to determine when the illness first manifested itself. This approach can also be used to identify the primary etiological component of theillness, which can be deduced from the study's findings. Our dataset was collectedfrom the “Popular Diagnostic Center – Savar branch”, and UCI databases. “RandomForest, Naive Bayes, Decision Tree, K-Nearest Neighbor (KNN), XGBoost, AdaBoost”, and many other complex and adaptable algorithms are requiredtooptimize the performance of this system. XGBoost was chosen as the most accuratealgorithm, with an accuracy of 99.1 %, according to the results. The overall performance of this method is excellent for both negative and positive values, as well as for the Macro and Weighted Average variables.