摘要
解释和评估模式是知识发现过程中的一个重要步骤。虽然在数据挖掘的算法中通过设置模式的重要性阚值可以消除大量无关模式,但当面对一个大的数据库时,数据挖掘的最终结果依然很大。从客观和主观两个方面分析了模式兴趣度的影响因子,并用多目标决策方法PROMETHEE对数据挖掘的结果进行综合评估,力图最终自动提交给用户的是易于理解的、潜在有用的有趣模式(知识)。最后通过实例说明了该方法的有效性和实用性。
It is an important step to interpret and evaluate the data mining results (patterns) in the process of knowledge discovery. Although large amounts of irrelevant patterns can be removed by setting a threshold as the pattern importance degree, the final results are still very great when a huge database is faced. This paper analyzes the influence factors of pattern interestingness in objective and subjective aspects, introduces a multi-object decision method to evaluate the results of data mining, so that the ultimate results are understandable, useful, and interesting to the end-user. Finally, an example shows the effectiveness and practicability of the method.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2003年第9期1090-1093,共4页
Systems Engineering and Electronics
基金
国家档案局攻关项目资助课题(2001-X-04)
关键词
数据挖掘
模式评估
兴趣度
多目标决策
Data mining
Pattern evaluation
Interestingness
Multi-object decision
