Predictive Machine Learning Approaches for Chronic Kidney Disease
Somya Srivastav, Kalpna Guleria, Shagun Sharma
Abstract
It might be challenging to diagnose chronic kidney disease (CKD) in its early stages due to the lack of symptoms. The creation and validation of a predictive model for the prognosis of CKD is part of the proposed work. Nowadays, it is becoming a common practice to predict and categorise diseases using machine learning algorithms. Inaccuracies and factual errors are common problems in medical records. In this work, the examination and performance improvement of the three machine learning classifiers, including Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), and Naive Bayes have been done. A CKD dataset has been used and collected from the UCI Machine Learning repository which contains 25 features. With the help of two classes from the CKD dataset, machine learning classifiers were created. After that, non-linear features and categories have been used to identify Kidney Disease. The results show that by using the random forest model, an average accuracy of 89.75% has been achieved, which is the highest among all the models taken for the study.