Applying rough set theory for filtering large number of coronary artery disease (CAD) rules

Abstract

This research is to apply Rough Set Theory (RST) method for filtering large number of generated rules from Cleveland Coronary Artery Disease (CAD) data set. Three stages are applied. First stage is rule extraction to get number of rules, second stage is rule filtering based on support value and the last stage is rule selection which is using Rough Set to reduct attribute. Every stage is being computed by using Rosetta software. Using validation data is to see the generalization improvement of RST method. The result of this experiment on Cleveland CAD data sets shows that RST method still has better accuracy than unfilterd rules based on accuracy on data testing which is 0.859504 and for validation data the accuracy is 0.770492 and there is no accuracy improvement from unfiltered rules but better than using other attribute selection.