Analysis of Machine Learning Algorithms in Predicting Chronic Kidney Diseases Based on Performance: A Systematic Literature Review

Chronic Kidney Disease (CKD) has been increasing nowadays from 750 million cases in 2019 to 850 million in 2023. Lack of access and late detection are one of the reasons why CKD cases have been increasing for all these years. Various Machine Learning (ML) algorithms have been studied to detect CKD and this paper will discuss and compare different types of algorithms to detect CKD early before progressing to advanced stages based on its performance. Although the use of ML to predict chronic kidney disease in the medical world is still hard to find, in other literature reviews with similar topic, it is found that various ML algorithms have been developed with the goal of assisting medical professionals in predicting CKD. Highly accurate ML algorithms have also been discovered using Artificial Neural Network, Random Forest, and XGBoost, despite needing further improvement in quality to be able to be applied on a large scale. In this paper, 5 commonly used ML algorithms namely XGBoost, Random Forest, Support Vector Machine (SVM), Artificial Neural Networks (ANN), and logistic regression were analyzed. Using Systematic Literature Review as this paper's methodology, this paper provides detailed explanation on how the algorithms work, implemented, and their performance were also analyzed by using 5 parameters: AUC, accuracy, precision, recall or sensitivity, and F1-Score. It was found that XGBoost achieved the highest score among all parameters except for AUC, which highest score was possessed by Artificial Neural Network.