A data reduction method based on indiscernibility and rough entropy for uncertain systems

Rough set theory is composed with a number of properties which makes it suitable for manipulating the uncertain systems. This uncertainty is handled in terms of lower and upper approximations which computes indiscernibility and rough entropy. In this paper a data reduction method based on indiscernibility and rough entropy is used. Firstly, the instance is compared to extract distinct instances on the basis of decision attribute. Secondly, on the basis of indiscernibility, the irrelevant attributes are removed. Thirdly, rough entropy is used to select the attributes based on their minimum entropy value. Finally, a significant amount of data is reduced using the proposed method. The effectiveness of the method is verified on eight standard medical data sets using classification accuracy. Experimental work demonstrates that with the reduced data SVM achieves a high classification accuracy.