Through repartition for the real domain interval values and the decision classifications, an extended method of
calculating the general important degree of an attribute and attribute subsets is proposed. A general neighborhood relation is
established on objects space. Then a rough set model is constructed based on the general neighborhood relation. Furthermore,
an approach for attribute reduction in a real domain decision system(ARRDDS) is developed. Results of experimental
evaluation on different data sets show that ARRDDS has the better performance comparing with the related rough sets
approaches of attribute reduction in processing decision tables with real-valued attributes.
Key words：general important degree；attribute reduction；neighborhood relation；real