An Integrated Framework for Effective Prediction of Chronic Kidney Disease

Kidneys play a vital role in the human body. They remove waste products and extra water from our bodies. It helps make red blood cells and balances important minerals. Chronic kidney disease happens when the Kidneys of the human body are damaged or feel difficulties while performing their tasks. Chronic Kidney disease can cause other diseases too like heart disease and stroke. That's why the prediction of chronic kidney disease is very important for avoiding these diseases. In previous work, only single algorithms were used which resulted in low accuracy. This paper analyzes the application of machine learning in the prediction of Chronic Kidney Disease, by examining several single classifiers such as Na¨ıve Bayes, Decision Tree, SVM, KNN, Random Forest, and ensemble methods. The dataset is taken from Kaggle's “Chronic Kidney Disease” which has 13 attributes and 400 samples. After filtering the dataset with no missing value to pass through, we applied the Synthetic Minority Oversampling Technique to overcome the issue of data imbalance. Then the cross-validation with f=5 is applied. By using single algorithms Random Forest, Support Vector Machine, Decision Tree, Na¨ıve Bayes, K-NN (k=3), we get the respective accuracies of 99.60%, 95.60%, 97.20%, 93.80% and 76.20%. We find a high accuracy of 99.60% with the random forest algorithm. To get better results we applied cross-validation (f=5) with an ensemble classification voting technique by a combination of multiple algorithms. We get the best results by the combination of decision tree, Naive Bayes, and random forest, which is 99.80%.