A New Algorithm of Discretization of Consecutive Attributes Based on the Decision in Rough Sets

Under the base on analyzing the traditional methods of the minimal decision rules in Rough Set Theory, defines the concept of the decision dependability and proposes a novel algorithm of obtaining the “optimal” decision rules as many as possible from the shortest combinations of condition attributes. The length of decision rules is necessary to be extended only when the current decision rules can’t represent all samples. At the same time, three methods are proposed to reduce the computational complexity: 1) defines the concept of bound coefficient, 2) only classify the samples with the same decision values at a time, 3) defines the Remain set. Above-mentioned methods can be used directly for incomplete information systems and have great practicability.