Predicting Chronic Kidney Disease of Diabetes Patients Using Ensemble Learning

Abstract

Chronic kidney disease is the reason for many deaths all over the world every year. Chronic kidney disease has troubled almost 753 million people all over the world in 2016, wherein 417 million are females and 336 million are males. In the year 2015, it was the reason for 1.2 million deaths all over the world. When CKD is detected in the later stage of a diabetes patient, it is very harmful to them. Sometimes it leads them to death. But if it is possible to detect chronic kidney disease at an early stage of diabetes patients, the damage can be minimized. This research paper has shown a comparative analysis on the performance of some algorithms - Multilayer Perceptron, Bagging, and Adaboost. And this research work has also used some algorithms such as Bagging (J48), Bagging (Random Tree), Bagging (Decision Stump), Bagging (LMT), Adaboost (Random Tree), Adaboost (Decision Stump), Adaboost (J48), Adaboost (Random Forest). Our comparison of different algorithms will help people having diabetes to figure out whether they will have CKD or not in the future. From all these algorithms Bagging (Random Tree) and AdaBoost (Random Forest) have the best result. By comparing the results of all algorithms, the best algorithm can be detected for predicting the chronic kidney disease. This study can save many people's lives and money. Doctors can also be benefitted from this research.