We point out that an absolute attribute reduct is generally not but contains a relative attribute reduct.

We study in this paper relations between absolute and relative attribute reducts of an information table.

The influence of the necessary attributes after class subdivision on the upper and lower approximate values,quality and accuracy of approximate classification,number of decision-making rules and relative reduction are analyzed theoretically.

By means of the methodology of rough set,different attribute classification methods for decision systems are first analyzed in this paper,and the attribute significance as well as the extremely small subset of the attribute relative reduction caused by a classification variety is discussed.

Two parameters for different level partitions (or similar degree between objects) α,β are used in these reductions, in which the positive region formula of the decision level set are adopted for relative reduction and importance degree of attributes, and the distributed reducti.

In this paper we investigate relations among absolute reduction and relative reductions,including reductions corresponding to objects and reductions corresponding to decision systems in inconsistent decision systems.

Therefor,the concept of absolute reduction in inconsistent decision table based on dominance relations is introduced,and it is proved that the absolute reduction is equal to the distribution reduction and maximum distribu-tion reduction.