Ensemble sparse intelligent mining techniques for diabetes diagnosis

Diabetes could be efficiently managed by discovering it early. If neither adequately treated nor predicted promptly, it could lead to various severe concerns like heart problems, kidney damage, and nerve damage. Out of these diabetes correlated ailments, chronic kidney disorder is the major one, and predicting it could ensure diabetic patients’ care. There is much correlation between occurrence of chronic kidney disorder and that of diabetes. The ability of the kidneys to filter waste items from the blood will deteriorate with diabetes. The presence of microalbuminuria and high protein levels in the urine indicates that the kidneys are not working correctly and the diabetic is having chronic kidney disorder. In this work, multi-layer perceptron, extra trees classifier, linear discriminant analysis, and stacking classifier are used to predict diabetic patients affected by chronic kidney disorder by utilizing historical data of diabetic patients. Finally, after evaluation of the considered methods using novel metrics, stacking ensemble classifier is identified as the better algorithm than their counterparts.