Prediction models for coronary heart disease

Abstract

In the current days, it is known that a great amount of effort is being applied to improving healthcare with the use of Artificial Intelligence technologies in order to assist healthcare professionals in the decision-making process. One of the most important field in healthcare diagnoses is the identification of Coronary Heart Disease since it has a high mortality rate worldwide. This disease occurs when the heart’s arteries are incapable of providing enough oxygen-rich blood to the heart. Thus, this study attempts to develop Data Mining models, using Machine Learning algorithms, capable of predicting, based on patients’ data, if a patient is at risk of developing any kind of Coronary Heart Disease within the next 10 years. To achieve this goal, the study was conducted by the CRISP-DM methodology and using the RapidMiner software. The best model was obtained using the Decision Tree algorithm and with Cross-Validation as the sampling method, obtaining an accuracy of 0.884, an AUC value of 0.942 and an F1-Score of 0.881.