Attention-Based Deep Neural Network for Coronary Heart Disease Risk Prediction

Tsatsral Amarbayasgalan, Kwang Ho Park, Lkhagvadorj Munkhdalai, Keun Ho Ryu

Abstract: Coronary heart disease (CHD) is one of the top causes of global mortality. It does not give any symptoms at the early stage, and most patients are diagnosed during periods of CHD exacerbation. If CHD reaches a hard condition, it will require advanced and long-term treatments. Moreover, the early detection of CHD is not easy; doctors make a decision based on many kinds of clinical tests. In recent years, machine learning-based approaches are widely used in the healthcare domain to predict disease. In this study, we have proposed a deep neural network (DNN) model with an attention layer that gives importance weight to the input of the output layer for the CHD risk prediction model. The proposed method has been evaluated by predicting CHD risk in the Korean population and compared with typical machine learning algorithms. As a result, the proposed attention-based DNN improved the performance of DNN and outperformed regular machine learning classifiers. The performance measurements include accuracy, precision, recall, f1-score, and AUC score reached 82.7818%, 86.6499%, 77.2704%, 81.6631%, and 82.5802%, respectively.

External IDs:doi:10.1007/978-981-33-6757-9_50