Dimensionality Reduction Using Neuro-Genetic Approach for Early Prediction of Coronary Heart Disease

This paper presents the development of a Neuro-genetic model for the prediction of coronary heart diseases. The novelty of this work is feature subset selection using multi-objective genetic algorithm without sacrificing the accuracy of ANN based heart disease predictor. Subsequently, the selected feature subset is used to predict the level of angiographic coronary heart disease using neural networks. The performance of the developed Neuro-Genetic model is evaluated using heart disease database obtained from Cleveland Clinic Foundation Database with all attributes are numeric-valued. The accuracy of the designed Neruo-Genetic model is validated using 303 patient data sets obtained for different age groups. This study exhibits early detection of heart disease with high testing accuracy of 89.58% through minimized feature subset, thereby reducing the complexity.