高效的時間管理是一個成功企業重要的特徴之一， 對大多數的企業而言，它也是一個必須時時精進與改善的議題。在企業每日營運活動中，異常流程會破壞整體的協同性與敏捷性，甚至會導致重大的損失。 優秀的管理者必須能不斷地找出潛在的異常流程，提前做出預防性的動作以避免發生不正確的流程。本研究採用 k-Nearest Neighbor 演算法，能夠計算出各流程的活動層級時間，並利用這些計算的結果，偵測出企業營運流程中異常的流程實例。 這些活動層級的時間，包括執行時間(execution), 傳遞時間(transmission), 等待時間(queuing) and 延遲時間(procrastination). 除此之外，流程的運行過程中，也會受到代理人如業務人員，生管人員，客戶及其他因素影響。這些影響流程運作的本文訊息(contextual information) ，能夠用模糊集合論的歸屬函數(membership functions)呈現出來，並用以校正活動層級的時間。 此演算法被植入一個全球性的中小企業的資訊系統中，經過一年的運作後，我們擷取該企業的資訊系統日誌並用以偵測異常流程. 實証結果經專家驗証，正確性達 81% ;Effective time management is one of the most crucial characteristics of a successful business. For most businesses, time management is an area that can always be improved. Irregularities in execution duration of business processes impede corporate agility and can incur severe consequences, such as project failure and financial loss. Efficient managers must constantly identify potential irregularities in process durations to foresee and avoid process glitches.
This paper proposes a k-nearest neighbor method for systematically detecting irregular process instances in a business by using a comprehensive set of activity-level durations, namely execution, transmission, queue, and procrastination durations. Moreover, because agents, customers, and other variables influence the progress of processes, contextual information is presented using fuzzy values. The values and corresponding membership functions are used to adjust the duration of each activity.
This proposed method was applied to the system logs of a medium-sized logistics company to identify irregularities. Experts confirmed that 81% of the instances identified as irregular were abnormal.
