應用基因流程探勘技術建構一個具有時間間隔的流程模型

Using Genetic Process Mining Technology to Construct a Time-interval Process Model

陳怡靜(I-Ching Chen)

指導教授 ： 蔡介元

元智大學/工程學院/工業工程與管理學系/碩士(2008年)

https://doi.org/10.6838/YZU.2008.00308

摘要

為了了解執行流程中事件與事件是如何進行，目前流程探勘技術已被廣泛的被使用。然而，現有大部分的流程探勘技術利用區域搜尋方法從歷史紀錄資料中逐步找出事件的關係而建立一個流程模型，使得事件關係對於整個歷史資料未能完整且正確的被表達出來。此外，有關於事件與事件間發生之時間紀錄並未被考慮於流程探勘技術中，使得不同時間間隔的事件會被視為相同的行為。另外，在目前的流程探勘技術中缺乏一個能夠評量流程模型品質的量測方法，以致於所得到流程模型的品質無法被斷定。為了解決上述在流程探勘技術中所欠缺的問題，本研究將利用基因演算法為基礎的流程探勘方法，透過族群大小(population size)、世代數(generation)、交配率(crossover rate)以及突變率(mutation rate)參數的變化找尋更佳之流程模型。同時，本研究將事件間發生之間隔時間納入建構流程模型的過程中，以幫助管理者觀察事件間不同時間的行為模式。最後，本研究亦發展一個流程模型品質之測量方法以評量所探勘出來的流程模型之品質，以協助管理者選取一個最佳的流程模型。

關鍵字

流程探勘 ； 基因演算法 ； 時間間隔

並列摘要

Nowadays, some process information is represented by a process model. To understand process executed in many activities, process mining technologies are now extensively studied to understand the relations and sequences between events (tasks). However, three major problems in the current process mining techniques are identified. First, the problems such as invisible tasks, noisy and non-free choice constructs and so on are difficult to be handled and current process mining techniques mainly based on the local search strategy which is to build the process model with information step by step. Second, time stamp are not considered so that the patterns with different time-intervals are regarded as the same behaviors. Third, there is no precision evaluation measure to evaluate the quality of process models in existing process mining techniques. To solve these difficulties, this research proposes a time-interval process mining method which considers time-interval between tasks. Furthermore, to solve the process mining techniques mainly based on the local search strategy to discover the process model and the discovered process model may contains the problems such as invisible tasks, noisy and non-free choice constructs, a genetic process mining method with global search strategy is applied. However, to evaluate the quality of the process models constructed by different combinations of parameters such as the population size, the number of generation, crossover rate and mutation rate in the genetic process mining method, a precision evaluation measure is proposed. Finally, a best process model with highest quality is selected to manage the real events (tasks).

並列關鍵字

Process Mining ； Genetic Algorithm ； Time-interval

參考文獻

7. Cook, J.E. & Wolf, A.L., "Discovering models of software processes from event-based data", ACM Transactions on Software Engineering and Methodology, vol. 7, no. 3, pp. 215-249, 1998.

10. Cook, J.E., Du, Z. & Wolf, A.L, "Discovering Models of Behavior for Concurrent Workflows", Computers in Industry, vol. 53, no.3, pp. 297-319, 2004.

13. Cook, J.E. & Wolf, A.L., "Event-based detection of concurrency", In Proceedings of the Sixth International Symposium on the Foundations of Software Engineering, pp. 35-45, 1998.

14. Chiola, G., Dutheillet, C., Franceschinis, G. & Haddad, S., "A symbolic reachability graph for coloured petri nets", Theoretical Computer Science, vol. 176, no. 1-2, pp. 39-65, 1997.

18. David, R. & Alla, H., "Petri nets for modeling of dynamic systems: A survey", Automatica, vol. 30, no. 2, pp. 175-202, 1994.