The Comparison of ReliefF and C.45 for Feature Selection on Heart Disease Classification Using Backpropagation

Abstract

One of the datasets used to predict heart disease is UCI dataset. unfortunately, the dataset contains missing data. the missing data dramatically affects the performance of the backpropagation classification method. One of the techniques used to handle missing data is feature selection. This study compares the ReliefF and the C4.5 algorithm in feature selection to handle missing data. The results of these algorithms are applied to the classification of heart disease using the Backpropagation. The results will be measured based on accuracy, precision, and recall. The performance results of the ReliefF and Backpropagation are an accuracy of 82.653%, a precision of 82.7%, and a recall of 82.7%. The performance results of the C4.5 and backpropagation are an accuracy of 80.61%, a precision of 80.4%, and a recall of 80.6%. Based on the results it can be concluded that the ReliefF gives better performance results on backpropagation than the performance results of the C4.5. Although, the results of C4.5 are below ReliefF but the results are quite satisfactory because of the accuracy, precision and recall results obtained above 80%. This shows that ReliefF and C4.5 can select features that affect the UCI heart disease patient dataset.