Information entropy based reduct searching algorithm

In this paper, a new reduct searching algorithm is proposed to benefit the applications of rough sets theory. The information entropy is introduced to the reduct searching algorithm, so that indeterministic causalities among attributes can be found and the reduct sensitivity to noise, occurring unavoidably when the approximation quality function γ is applied, can be removed. The theoretical analysis and an illustrated example show that the rule set induced by this algorithm uses less attributes than that by algorithms based on the rough set γ function. At the same time, the new algorithm gives a larger rule set coverage, especially when the data are polluted by noise or the causalities among the attributes are indeterministic. In practice, noise pollution of data and indeterministic causalities are usual situations, so this algorithm is more applicable than those based on the rough sets γ function.