基于含噪音日志的流程挖掘阈值优化设置  

Optimal Setting of the Threshold in Mining Process Model from Noised Log

作　　者：阮莹 苏强[1,2,3] 张国通[4] 刘大庆[4] 戴红芳[4] 张音赟[4] 朱岩[3] 薛镭[3] 

机构地区：[1]上海交通大学工业工程与物流工程系,上海200240 [2]同济大学经济与管理学院,上海200092 [3]清华大学经济管理学院,北京100084 [4]上海市第七人民医院,上海200137

出　　处：《上海交通大学学报》2010年第2期276-281,共6页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金资助项目(70672077,70832006);上海交通大学医工(理)交叉研究基金重点项目(YG2007ZD10);清华大学经济管理学院医疗健康管理研究中心资助项目

摘　　要：针对处理日志中噪音数据的启发式流程挖掘算法中阈值设置的不确定性,提出了基于试验设计的阈值优化设置方法.以阈值作为变量,挖掘得到的流程模型与实际日志的符合度作为响应量,通过试验设计分析方法优化阈值的配置,并将该算法应用于医院某病种诊疗流程的挖掘.结果表明,通过该阈值设置方法能挖掘出正确合理的流程模型.In view of the uncertainty of the settlement of the threshold in the heuristic process mining method proposed by Aalst to deal with the noise data in the log, a method of optimization settlement of threshold based on design of experiment （DOE） analysis was proposed. The threshold is dealt as variable, and the fitness of the model that is mined as response variable, the goal is to find the most optimal combination of threshold value that will result in the most appropriate workflow model. Finally, this method was applied to mine Caesarean birth diagnosis flow. The result demonstrates that this method can find an optimal combination of threshold that result in an appropriate workflow model.

关 键 词：流程挖掘 噪音数据 阅值 插值法 试验设计 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]