The method of attribute reduction based on discernibility matrix

With respect to decision table, especially the inconsistent ones. On the basis of equivalent partitions, This paper quantifies the elements in the same class deduced by condition attribute belonging to different decision attribute classes, generates the discernibility matrix and measures the indexes of attribute significance reasonable, and then uses these formulas as heuristic information to design a heuristic algorithm for attribute reduction. Moreover, a process of reducing again is used in it in order to eliminate redundancy attribute. Finally, an example demonstates the attribute reduction algorithm can be used in decision tables effectively.