An approach to knowledge reduction based on relative partition granularity

Knowledge and classifications are related together by the theory of rough sets which claim is that knowledge is deep-seated in the classification abilities of human beings. In this paper, relative partition granularity, a quantitative representation for the relative classification ability of conditional attributes relative to decision attribute was defined. The equivalence between some basic concepts in rough set theory and relative partition granularity was proved. A heuristic knowledge reduction algorithm was designed based on relative partition granularity. Finally, we show that this algorithm is effective through an example.