一种基于多视角数据属性的决策表Petri网挖掘方法

当处理高度可变的流程时，已有的自动过程挖掘技术产生的模型可能并不能真实反映流程运行中不同决策点之间规则的变化情况。从声明性过程挖掘的角度出发，提出了一种具备可视化规则的决策表Petri网挖掘方法，实现真实日志到声明性过程决策表Petri网模型的映射。首先，形式化了决策表Petri网模型及其携带的规则分析决策表，并对模型的静态语义和动态语义进行定义；其次，通过扩展属性的添加，分析流程内部属性和事件属性是否会对决策产生影响，并通过规则分析决策表的异常值属性，判断规则的异常程度；最后，在一组人工日志和真实事件日志的基础上进行实验仿真，并与数据Petri网的挖掘技术进行分析对比。实验结果表明所提方法在反映流程运行中规则的变化情况具有一定优势，并为数据流异常检测提供数值可解释性；同时，所设计的决策表Petri网挖掘方法可以将决策信息与模型结构整合在一起，为过程模型的可变性建模提供形式化基础。

Approach of decision table Petri net mining method based on multi perspective data attributes

When dealing with highly variable processes, the models may not accurately reflect the changes in the rules between different decision points in the process operation, by using the existing automatic process mining techniques. From the perspective of declarative process discovery, this paper proposed a decision table Petri net mining method with visual decision rules, to realize the mapping from real event log to a declarative process decision table Petri net model. Firstly, this method formalized both the decision table Petri net model and its rule analysis decision table, and desined the static and dynamic semantics of the formalized decision Petri net model. Secondly, through adding extended attributes, it analyzed whether the internal attributes or event attributes of the process would affect the decision. Furthermore, it generated outlier attributes with their deviation degrees of the rule analysis decision table to determine the degree of exception of the rule. Finally, it conducted experimental simulation on the basis of a set of artificial logs and practical event logs, and at the same time, it analyzed the experimental results by comparing with the extant data mining technology by Petri nets. The experimental results show that the proposed method has certain advantages in representing the change of rules during the operation of the process, which can provide quantitative and interpretable analysis for data flow anomaly detection. At the same time, the proposed decision table Petri net mining method can integrate the decision information with the model structure together, providing a formal basis for the variability modeling of the process model.