In Exchange 5.5 (and earlier), a site defined both the administrative boundary and the physical routing topology for a group of servers. 在Exchange 5.
This method chooses the boundary region of rough sets as the criterion of selecting partitional attributes and redefines the conception of confidence of leaf nodes. 经实验证明，该方法能有效提高分类效率且更加容易理解。
Secondly, misclassifications in the boundary region between the two classes are in fact unavoidable due to the complexity of TSA input dimension and the limitation of ANN. 其二,由于暂态稳定评估问题输入空间的复杂性以及神经网络本身的局限性,神经网络的分类结果中不可避免地会存在误分类。