Diagnosis of heart disease using a committee machine neural network
Heart disease is one of the leading causes of deaths worldwide. Several methods have been developed by researchers to support medical diagnosis of heart disease, including artificial intelligence methods. In the past, committee machines have been shown to achieve higher classification accuracy than a single classifier. This study uses a committee of classifiers consisting of a combination of feed-forward multi layer perceptron (MLP) and radial basis functions neural networks (RBF) to diagnose heart disease. The output of the committee has been obtained based on majority voting. Several MLP training algorithms have been analyzed from the viewpoint of learning performance based on the network topology to find the network with the best prediction results. Cleveland heart disease dataset has been used throughout the experiments. The results show that the committee machine approach gives significantly better results than a single neural network. The classification accuracy obtained by the proposed method achieves a high accuracy rate of 95,4545%. This result is better than the results achieved by other methods that use Cleveland dataset reported in literature to this date.