摘要
为了能表示数据的层次结构,Wu和Leung从粒计算的角度提出了一种新的数据表示方法称为多尺度决策表,并提出了最优尺度的定义,在最优尺度上进行属性约简、规则获取等任务可以提高知识发现效率.但是Wu和Leung所提出的基于协调性的最优尺度选择的方法在面对不协调的多尺度决策表的时候,仅考虑广义决策而忽视了不协调的多尺度决策表决策规则置信度的变化.针对这个问题,本文提出了基于熵的最优尺度选择的方法,该方法无需判断多尺度决策表的协调性,就可以获取保持决策规则置信度的最优尺度.
In order to represent the hierarchical structure of data,Wu and Leung proposed from the perspective of granular computing a new data representation model,namely,the multi-scale information table and the multi-scale decision table. Meanwhile,they proposed the definition of the optimal scale,on which the tasks performed,such as attribute reduction and rule acquisition,can improve the knowledge discovery efficiency. However,when it comes to the inconsistent multi-scale decision table,the method of optimal scale selection based on consistence as proposed by Wu and Leung,focuses solely on the consistence of the generalized decision values at the optimal scale,while ignoring the changes of confidence about the decision rules in the inconsistent multi-scale decision table and missing some information with comparison to the original data. To solve this problem,this paper comes up with a method of the optimal scale selection based on entropy,taking into consideration the changes of confidence about the decision rules without judging the consistence of the multi-scale decision tables. In summary,the current method is more convenient and effective than that proposed by Wu and Leung.
作者
魏巍
张嘉宇
陈千
崔军彪
WEI Wei;ZHANG Jla-yu;CHEN Qian;CUI Jun-biao(a.School of Computer ＆ hfformation Technolog;b.Key Laboratory of Computation Intelligence ＆ Chinese Information Processing,Ministry of Education,Shanxi University,Taiyuan 030006,China)
出处
《海南热带海洋学院学报》
2018年第5期61-65,共5页
Journal of Hainan Tropical Ocean University
基金
国家自然科学基金项目(61303008
61502288
61403238)
山西省自然科学基金项目(201701D221101)
上海市科学技术委员会科研计划项目(16JC1403000)
关键词
信息熵
多尺度决策表
最优尺度选择
information entropy
multi-scale decision table
optimal scale selection
