Rough set theory is an important tool to deal with uncertain or vague knowledge. In this paper, the Rough set theory is deeply
investigated, and an approach for data filtering based on rough set theory is proposed. The important feature of this approach
is that the internal dependency structure of the system is kept intact, and that no additional parameters are needed. Theoretical
analysis and experimental results show this approach can effectively reduce granularity of attribute measurement and improve
the statistical signification of rules.