ROUGH SET THEORY BASED REDUCTION ALGORITHM FOR DECISION TABLE

With the purpose to reduce the surplus information on decision table and extract the determinative rules, an autonomous clustering algorithm based on graded datum subtraction (ACGDS) is proposed to reduce the data area and an attribute reduction algorithm based on ant colony optimization (ARACO) is presented to reduce the surplus attributes. ACGDS uses the quick sort method and subtraction to every row of the similarity matrix only depending on data attributes. ARACO directly imports the core into the distributing of initial pheromone, and reduces the problem scale, and solves the low convergence speed problem in the conventional ant colony algorithm. The experiments illustrates that ACGDS increases the accuracy and efficiency and ARACO could find the minimal reduction with less time. It is worth to use rough set theory to deal with the problem of decision table reduction.