Predict the Diagnosis of Heart Disease Using Feature Selection and k-Nearest Neighbor Algorithm

In this paper, the prediction of heart disease based on feature selection by using multilayer perceptron with back-propagation algorithm and k-nearest neighbor algorithm based on an explicit similarity measure with biomedical test values to diagnose heart disease is presented. The main motivation for this paper is to classify the heart disease with reduced number of attributes. We use the weight information by a multilayer perceptron to determine the attributes which reduces the number of attributes which is needed to be taken from original datasets (13 attribute is reduced to 8 attributes). Afterward, we used k-nearest neighbor algorithm to predict the diagnosis of heart disease after the reduction of a number of attributes. The accuracy differs between 13 attributes and 8 attributes in testing data set is 93% and 90%, respectively. The experimental results show that our propose classification help in the best prediction of heart disease which even helps doctors in their diagnosis decisions.