顧客對產品的差異性與多樣性的要求日益增加,因此企業盡可能地將造成產品差異性的作業往後延,這樣能降低不確定與風險,且能快速回應顧客需求,於是產生了延遲(postponement)策略的概念。在執行工作流程管理系統中,可以挖掘工作流程中事件日誌(event logs),於是就產生工作流程日誌(Workflow log)的相關研究。我們可藉由workflow log來挖掘出新的延遲策略,並建議新的延遲策略來改善公司的流程。 本研究目的在於使用工作流程探勘技術,挖掘出工作點的產出資料與決策點的產生結果,來分析彼此之間的關係,如果決策點的產出結果,不會影響工作點,則可以將決策點往後延遲。本研究提出決策點延遲演算法,可以自動的將工作流程圖中的所有決策點進行延遲的動作,將會得到新的工作流程圖,此流程圖中決策點都會是最延遲的情況,且符合完整性與正確性。 In order to meet increasing customer demands for more diverse product offerings, firms are delaying the delivery of the diverse products. This will reduce risk and uncertainty and speed up the rapid response to customer needs, termed Postponement Strategy. The extraction of workflow models from business processing logs is called Workflow Mining. We can implement Postponement Strategy with workflow logs and recommend new Postponement Strategies to improve the process. Our goal is analyzing the dependencies of the decisions made at the decision point and the data produced by the tasks of the process. If the output at the decision points does not influence the tasks, the decision points can be postponed. We develop a Decision Point Postponement Algorithm for auto-postponing all the decision points and finally transforming a new workflow graph. All decision points of workflow graph will be the latest condition and correspond with Completeness and Correctness.