Medical diagnostic rules as upper approximation of rough sets
Shusaku Tsumoto · 2002
Rule induction methods have been proposed in order to acquire knowledge automatically from databases and applied to data mining applications. However, conventional approaches focus on the positive aspect of diagnostic rules, which is corresponding to the lower approximation of a target concept in rough set theory. However, an original data set include the negative aspects of the data set, which can be derived from the upper approximation of the target concept. In this paper, medical diagnostic rules are defined from the upper approximation and a new rule induction algorithm is defined by using these definitions.