Abstract: By investigating the concrete instances, this paper illuminates classification qualities, relative positive region, non-monotonic decreasing characteristics of classical rough set model and the occurring bounce phenomena of classification qualities and relative positive region in the process of reduction for variable precision coverage rough set model as well．However, there maybe some instabilities in the procedure above. So it is necessary to establish different models according to them in order to make the reduction of properties in variable ways.

Key words: Pawlak Z rough set, variable precision rough set, non-monotonically decreasing, bounce phenomena

摘要： 通过抽象信息系统，阐明了经典粗糙集模型分类质量、相对正域、决策类下近似具有非单调递减性；变精度粗糙集模型在约简过程中分类质量和相对正域会出现跳跃现象，约简过程具有不稳定性。需要针对三者分别建立模型，使属性约简变得多样化。

关键词: 经典粗糙集, 变精度粗糙集, 非单调递减, 跳跃现象