Automated Chronic Kidney Disease Detection Model with Knearest Neighbor

Abstract

Chronic kidney disease is one of the most common disease in the world today. Kidney disease causes death if the patient is not threated at early stage. One of the challenge in kidney disease treatment is accurate identification of kidney disease at an early stage. Moreover, detecting kidney disease requires experienced nephrologist. However, in developing nations lack of medical specialist or nephrologist for identifying chronic kidney disease makes the problem more challenging. As alternative solution to kidney disease identification, researchers have developed many intelligent models using K-nearest Neighbors (KNN) algorithm. However, the accuracy of the existing KNN model has scope for improvement. Thus, this study proposed KNN based model for accurate identification of kidney disease at early stage. To develop optimized KNN model, we have employed error plot to find most favorable K value to obtain more accurate result than the existing models. To conduct experiments, study employed kidney disease dataset collected form publically available Kaggle data repository for training and testing the proposed model. Finally, we have evaluated the proposed model against predictive accuracy. The experimental result on the proposed model appears to prove that the predictive accuracy of the model is 99.86%.