Abstract:
CKD is a long-term disease that impacts a lot of people all over the world, including in Bangladesh, and results in high illness and death rates. CKD is an illness that peaks in a slow reduction day by day. Early detection is vital since timely treatment may halt the advancement of CKD, improve the quality of patients lives, reduce medical costs, and prevent the danger of future health issues. In the past ten years machine learning models have appeared as transformative tools in medical testing, Utlizing large data and difficult algorithms to detect symptoms unknown by the doctor. This research project implemented and accessed several machine learning techniques for detecting chronic kidney disease early. Targeting on their respective efficiency, assets, and limitations. Machine learning model converts health assessment by applying large data and cutting-edge algorithms to detect structures that possibly alternatively bypass healthcare professionals. This research project used the dataset from the UC Irvine Machine Learning Repository created by the University of California in 1987, and this dataset is publicly available in their website. In this research evaluator evaluated and differentiated the performance of Decision Trees, Logistic Refression, and XGboost. All the model are using k-fold cross validation for spiting the data into training and test part. A stacking model also implemented in this research with the combination of Decision Trees, Logistic Regression and XGBoost. Analysis result showed that stacking model named CKDML762 had the best performance generating a perfect accuracy, precision, recall, f1 score. This outcome indicates that the stacking model CKDML762 correctly stored all the cases in training and testing sets. Integrating machine learning algorithms into the CKD holds great potential changing clinical work. Medical experts can imrpove testing accuracy and facilitate timely measures by utilizing proposed algorithm such as CKDML762
