Families of the granules for association rules and their properties
Families of the granules for association rules and their properties
复制标题
关联规则的颗粒族及其属性
DOI：
10.1007/978-3-319-25754-9_16
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
M. Nakata
中科院分区：
文献类型：
--
作者：
H. Sakai;C. Liu;M. Nakata
我们使用了表中描述符定义的颗粒（或等价类），并研究了基于粗糙集的规则生成。在本文中，我们考虑了由蕴涵定义的新颗粒，并在具有精确数据的表中提出了由蕴涵定义的颗粒族。每个家族由四个颗粒组成，我们表明，使用这四个颗粒可以很容易地获得三个标准值，支持度，准确性和覆盖率。然后，我们将此框架扩展到具有非确定性数据的表。在这种情况下，每个家族由九个颗粒组成，并且通过使用九个颗粒也获得了三个标准的最小值和最大值。我们证明有一个表导致支持和精度最小，通常没有表导致支持，精度和覆盖率最小。最后，我们考虑了这些属性在基于apriori的不确定数据规则生成中的应用。这些属性将使基于apriori的规则生成更加有效。
We employed the granule (or the equivalence class) defined by a descriptor in tables, and investigated rough set-based rule generation. In this paper, we consider the new granules defined by an implication, and propose afamily of the granules defined by an implicationin a table with exact data. Each family consists of the four granules, and we show that three criterion values,support,accuracy, andcoverage, can easily be obtained by using the four granules. Then, we extend this framework to tables with non-deterministic data. In this case, each family consists of the nine granules, and the minimum and the maximum values of three criteria are also obtained by using the nine granules. We prove that there is a table causingsupportandaccuracythe minimum, and generally there is no table causingsupport,accuracy, andcoveragethe minimum. Finally, we consider the application of these properties toApriori-based rule generation from uncertain data. These properties will makeApriori-based rule generation more effective.