Performance Evaluation Software Engineering Laboratory,
Sharif University of Technology,
under supervision of Prof. J. Habibi.
Tehran, Iran, 2012 - 2013
Coronary Artery Disease (CAD) is the most common fatal heart disease. Present methods for CAD diagnosis are costly, time-consuming, and hazardous. Therefore, study of non-invasive methods has proved to be effective. We have worked on various techniques relying on advanced data mining methods of hundreds of patients’ medical information.
We sought effective features in CAD diagnosis other than those previously studied. Not only did we consider these new features, but we also proposed an ensemble algorithm, therefore higher accuracy was achieved. For improved diagnosis, we decided to examine the stenosis of each vessel separately by means of data mining algorithms and measure their accuracy with tenfold cross validation. We further advanced our algorithms by considering diagnosis of diseased patients vital, since misclassification of them has more side effects. For considering this fact, we proposed a new algorithm by use of various cost-sensitive algorithms.