A Machine Learning Perspective for Predicting Chronic Kidney Disease
B. Vanathi, S. Ramesh, K. Sudha, K Tamizharasu., N. Sengottaiyan, P Kalyanasundaram
Abstract
Aim: The primary objective of this research is to increase accuracy in the prediction of chronic kidney disease (CKD) by using Machine Learning (ML) algorithms, including K-Nearest Neighbors, Support Vector Machines, and Artificial Neural Networks algorithm. Methods and Materials: The proposed work included four groups. Group 1 refers to a set of four different ensemble tree ML algorithms (Random Forest, Extra Trees, AdaBoost, and XGBoost) that were used to obtain the optimal classification model to support CKD early diagnosis; Group 2 refers to the K-Nearest Neighbors algorithm, which can be used to handle missing values; Group 3 uses the Support Vector Machine algorithm to classify patients into CKD or non-CKD categories; and Group 4 refers to the Artificial Neural Networks algorithm that analyses medical data to predict CKD. Results: The proposed system improves chronic kidney disease prediction, achieving 99.2% accuracy for early detection and management on an automated platform. Conclusion: All three models, including KNN, SVM and ANN, have demonstrated their potential in accurately predicting CKD with an average accuracy of 99.2%, which performs better than four different ensemble tree ML algorithms.