Prediction of Chronic Kidney Disease Using Data Mining Techniques

Abstract

Recently renal failure disease has spread widely all over the world, especially in Sudan, as indicated by the WHO reports. Therefore, it was necessary to use all available scientific methods to contribute in studying the factors that lead to the disease and predict it in its early stage, to decrease its wide spread. In this research, data mining techniques were used to study and determine the factors that lead to Chronic Kidney Disease in its early stages, and to build models to predict the disease using the selected features. Data used in this research was collected from a Medical Center for Renal Failure Treatment in India. WEKA machine learning software was used in this research for all data mining operations like data exploration, feature selection, and model development. Supervised machine learning algorithms, such as Naïve Bayes, Random Forest, C4.5 Tree and Neural Networks, were used to select the important features and develop the models. Several models were built using several algorithms, each of which gave high accuracy and acceptable interpretation to the physicians. The research motivates other researchers to start working intensively in this field by forming research groups from data scientists and physicians to solve such problems using real patients’ data.