A comprehensive evaluation and meta-analysis of machine learning techniques for identifying and classifying chronic renal disease

Abstract— Heart failure (HF), atherosclerotic cardiovascular disease, and end-stage kidney disease are all signs of chronic kidney disease brought on by type 2 diabetes (CKD). However, because it has no symptoms, CKD is frequently overlooked or misdiagnosed. A person can only survive for around 18 days without kidneys, necessitating dialysis and a kidney transplant. It is crucial to have trustworthy techniques for early detection of CKD. Additionally, patients frequently forego the common urine protein-based CKD detection test. It describes a medical condition that damages the kidneys and has an impact on the body as a whole. Inadequate diagnosis and care can lead to end-stage renal disease, which eventually kills the patient. In response to the shortcomings of conventional biomarkers and the requirement for early therapeutic intervention in cats with CKD to improve outcomes, new renal biomarkers for the detection of glomerular or tubular dysfunction have been discovered and validated. Changes in the concentrations of these biomarkers in the blood or urine may reveal early kidney damage or forecast the development of kidney disease before changes in conventional biomarkers can be seen. CKD forecasting is effortless with machine learning (ML) approaches. The current study's strategy for predicting CKD status using clinical data involves data preparation, a mechanism for managing missing values, data aggregation, and feature extraction. We examine some recent works on ML-based CKD detection and classification in this systematic review. Data source, data preprocessing, feature extraction, feature selection, detection, and classification are the various tasks that make up the process of CKD detection and classification. To be more precise, we group this systematic review into two main categories: supervised learning, and unsupervised learning-based CKD detection and classification. Peer-reviewed papers were gathered using reputable search engines like Elsevier, IEEE Xplore, PubMed, Scopus, Web of Science, and others. Finally, we briefly discuss current problems and the requirements for linking academic research on CKD detection and classification to industry-based solutions that will improve standard clinical care.