论文部分内容阅读
利用粗糙集中的三支决策思想,将类用正域、负域和边界域刻画,得到初始聚类结果。然后通过定义重叠度和类与类的合并策略,将初始聚类结果进行合并,得到最终聚类结果。之后应用2个关系网络数据展示了具体的聚类步骤,并通过比较2个例子的聚类结果,分析了影响聚类结果的一个主要因素:阈值的选取。实验表明:阈值的选取对简单的网络结构数据集的聚类结果的影响并不明显,然而对复杂的网络结构数据集的聚类结果的影响则较为显著。
By using the three decision-making ideas in rough set, the class is described by positive and negative fields and boundary regions to get the initial clustering result. Then the initial clustering results are merged by defining the overlapping degree and the merging strategy of the class and the class to get the final clustering result. After that, we use the two relational network data to show the concrete clustering steps. By comparing the clustering results of two examples, we analyze one of the main factors affecting the clustering result: the threshold selection. The experimental results show that the influence of the threshold selection on the clustering results of simple network structure data sets is not obvious, but the impact on the clustering results of complex network structure data sets is significant.