PERFORMANCE OF DATAMINING TECHNIQUES IN THE PREDICTION OF CHRONIC KIDNEY DISEASE

Data mining being an experimental science is very important especially in the health sector where we have large volumes of data. Since data mining is an experimental science, getting accurate predictions could be tasking. Getting maximum accuracy of each classifier is necessary, it is therefore important that the appropriate feature selection method is selected. Feature selection is highly relevant in predictive analysis and should not be overlooked. It helps reduce the execution time and provide a more accurate and reliable result. Therefore more research on predictive analysis needs to be delved into and how reliable these predictions are.  Application of data mining techniques in the health sector ensures that the right treatment is given to patients. This study was implemented using WEKA.

This study is aimed at using 3 classifiers: multilayer perceptron, naive bayes and J48 decision tree in the prediction of chronic kidney disease dataset. The aim of this research work is to evaluate the performance of the classifiers used based on the following metrics- accuracy, specificity, sensitivity, error rate and precision. Based on the performance metrics mentioned above, results shows that J48 decision tree gave the best result but naive bayes had the lowest execution time therefore making it the fastest classifier.