Decision rule visualization for knowledge discovery by means of rough set approach

Recently, the application studies of rough set theory to real problems have been increasing. The discovery of useful decision rules is a key to a success of the rough set applications. However, finding useful decision rules is often difficult when a lot of decision rules are induced by the rough set approach. In this study, we develop a decision rule visualization system for supporting the discovery of useful decision rules. The system visualizes decision rules having same conclusions in three-dimensional coordinate having the vertical axis of co-occurrence rates between atomic formulae and conclusions and two other axes obtained by Hayashi´s quantification method IV using co-occurrence rates between atomic formulae as similarities. An experiment is conducted to evaluate the proposed system. The results show that useful decision rules are discovered efficiently from many decision rules. Moreover, we show the usability of the system by numerical experiments using car data.