Feature Selection by mRMR Method for Heart Disease Diagnosis

Heart disease has become a non-ignorable threat to human health in recent years. Once without timely diagnosis and treatment, patients often suffer disability or even death. However, the accuracy directly relies on different doctors’ experiences various factors associated with heart bring heavy tasks them make situation worse. Therefore, improve introducing computer-aided techniques assist doctors is feasible approach. At present, researchers usually use processed dataset (13 features) selected by from unprocessed (74 (UCI Machine Learning Repository) apply feature selection method dataset, it’s inappropriate because scale so small. People neglect dataset’s value don’t realize it could contain some latent information. A comprehensive comparison needed demonstrate advantages. Besides, incremental combination should be verified. As minimum Redundancy - Maximum Relevance (mRMR) gains great success selection, applying as filter can enhance classification accuracy. Thus, this research, we introduced mRMR for made within several methods like Principal Component Analysis (PCA), Linear Discriminant (LDA), Kendall, Random Forest, other research works metrics. By analyzing results, most cases, algorithm’s performance. The effective superior methods. Not only does own highest accuracies, but also least supportive features. It 100% 8 features Cleveland 98.3% 14 Hungarian, 99% 9 Long-beach-VA, respectively. Furthermore, find that features, which regard useless, play part classification, attract attention doctors.