基于统计偏好的边界域重构方法

Approach of Reconstruction about Boundary Region Based on Statistics Strategy Preferences

目前许多挖掘算法都试图使异常信息的影响最小化,或者排除它们,经典粗糙集理论基于正域的属性约简方法也不例外,它直接排除了边界域中样本所包含的信息.如何改变边界域结构,将边界域样本尽可能拓展到正域结构中,从而有效获取更有价值信息的研究很有必要.在经典粗糙集理论的基础上,采用统计学中基于某种偏好策略,提出了边界域重构的基本方法和知识获取方法,进一步讨论了与变精度粗糙集模型之间的联系,并重新定义了变精度粗糙模型中关于β下近似的定义.结果表明基于边界域重构的方法和修正后变精度粗糙集模型正域结构得到相应扩大,获取异常信息能力进一步加强.

At present, many mining algorithms try to minimize the impact of abnormal information, or even to eliminate them. The attribute reduction methods are not exceptional based on the positive region of classical rough sets theory, which directly eliminate the valuable information contained in the samples of the boundary region. It is necessary to study how to obtain more valuable information from the boundary region. Therefore, how to change the boundary region structure and expand into positive region samples as much as possible are necessary. Based on the classical rough sets theory, the paper puts forward the basic reconstruction about the boundary region and knowledge acquisition methods using preferences strategy of statistics. Based on it, the relationships with the variable pre- cision rough sets theory are discussed and definition of β lower approximation is redefined further. The results show that the positive regions are enlarging by the reconstruction method based on boundary region and the modified low approximate of variable precision rough sets theory, which can strengthen the ability of obtaining the abnormal information.