工作流挖掘：相关问题和方法的研究(13)

11. 结论

通过本文，我们提出了对与工作流挖掘相关的各种问题、技术、工具和方法的一些观点。了解第5-9章所讲述的5种方法的不同以及它们所面对的不同问题是一件很有趣的事情。较为正式的方法是在第5章中提到的使用Petri网理论来描述能被挖掘的工作流模型的类别。

第6章和第7章提到的启发性方法关注的是噪音和决定挖掘结果质量的因素等问题。与其他方法不同，第8章提到的方法关注的是具有同一标签（名称）的多重任务。最后，第9章提到了挖掘包含许多过程的块结构（即：结合了与/异或分支和与/异或连接）的方法。

第10章从9个方面（结构、时间、基本平行、非自由选择、基本循环、任意循环、隐藏任务、重复任务和噪音）对比了前面提到的方法。通过对比展示了方法间存在的不同，也指出了需要解决的问题。

为了团结力量、共享知识和发展成果，我们介绍了一种与工具无关的XML格式。这一格式在第4章中给出，我们希望能鼓励在这一领域的其他研究者/开发者使用这一格式。

11. Conclusion

In this paper, we presented an overview of the various problems, techniques, tools, and approaches for workflow mining. It is quite interesting to see how the five approaches presented in Sections 5–9 differ and are driven by different problems. The more formal approach described in Section 5 uses Petri-net theory to characterize the class of workflow models that can be mined.

The more heuristic approaches in Sections 6 and 7 focus on issues such as noise and determining the quality of mining result. Unlike the other approaches, the approach in Section 8 takes into account the fact that there may be multiple tasks having the same label. Finally, the approach in Section 9 exploits the block structure (i.e., corresponding AND/XOR splits and AND/XOR joins) of many processes. Each of these approaches has its strengths and weaknesses.

Section 10 compared the approaches by focusing on nine aspects (Structure, Time, Basic parallelism, Non-free choice, Basic loops, Arbitrary loops, Hidden tasks, Duplicate tasks, and Noise). This comparison reveals differences and also points out problems that need to be tackled.

To join forces and to share knowledge and development efforts, we introduced a tool-independent XML format. This format was given in Section 4 and we would like to encourage other researchers/developers in this domain to use this format.