关键词: 粗糙集, 决策树, 数据离散化, 数据挖掘, 谱系聚类, 属性约简

Abstract: Rough sets and decision tree have complementary characteristics. A new approach to data mining is thus proposed combining both advantages. Taking the detected data of plywood defects as example, the defects are recognized as follow using eigen information in the database of plywood on the basis of rough sets theory. Decentralizes the data in the database by the algorithm of center-of-gravity distance of pedigree cluster, then reduces the conditional attribute by use of rough sets to obtain the low dimensional sample data. Decision rules are finally obtained by decision tree. The experimental result shows that, in this way, the original characteristics of data remained unchanged, and the knowledge acquisition process become speedier so as to improve the classification accuracy of model and interpretability of rules. Comparing with other the methods, such as rough sets or precision-varied rough sets, the method is proved more satisfactory.