A Risk Assessment Model for Patients Suffering from Coronary Heart Disease Using a Novel Feature Selection Algorithm and Learning Classifiers

Abstract

The aim of this research is to develop an efficient risk assessment model to assess the risk in patients suffering from coronary heart disease. The proposed technique classifies patients as having low risk or high risk of coronary heart disease. CVD dataset from Cleveland database is used to develop the model. The various parameters considered are cholesterol, blood pressure, electrocardiogram and echocardiogram tests among others as well. The patients suffering from coronary heart diseases are labeled as lower or higher risk. Feature selection is one of the critical tasks in developing predictive models. It reduces the computational cost by removing insignificant features. This leads to a simpler, accurate and comprehensible model. Here, two feature selection techniques namely Mean selection (MS) and a novel feature selection technique, VAS-CHD, a variance-based attribute selection for coronary heart disease diagnosis is implemented to capture the right features for the purpose of prediction. The decision tree has resulted in accuracy of 80.5% and lower False Negatives for features obtained from MS. The multilayer perceptron has resulted in 73.7% accuracy and lower False Negatives for features obtained from VAS-CHD.