Dynamic Reduct Research based on Rough Set Theory

In this paper we propose a new dynamic reduct algorithm based on rough sets theory. When the number of the object in the information table increase, instead of treating the changed information table as a new one and finding the reduct again like rough set reduct algorithm does, the dynamic reduct algorithm just update the old reduct set based on the increased objects, so the computation time is greatly saved. Also the entropy criterion is introduced to the dynamic reduct algorithm, so the statistical causality between the attributes could be find and the reduct′s sensitivity to the noise unavoidably when the \$γ\$ criterion is applied could be removed. An illustrated example shows that by the dynamic reduct algorithm, computation time is greatly saved compared with the rough set theory based reduct algorithm, at the same time the same reduct is find with less objects, furthermore the results are sorted according to their causal strength.