A New Cube-based Algorithm for Computing the Feature Core of a Consistent Decision Table

This paper focused on computing the feature core of rough set theory by making full use of the aggregation information in a data cube. After we established a one-to-one mapping relation between equivalence classes in a decision table and nonempty cells in a data cube, a new cube-based algorithm for computing the feature core of a consistent decision table was put forward in this paper. The correctness of the new approach was proved. The algorithm is different from general methods for computing the feature core. It does not have to generate the discernibility matrix. The experiments with UCI data set show that the new approach has high time performance with small feature set and large data set.