應用資料探勘建構物流中心訂單需求模式

Applying Data Mining Techniques to Model the Order Demand of Distribution Center

侯琬伊(Wan-Yi Hou)

指導教授 ： 陳穆臻

國立臺北科技大學/管理學院/商業自動化與管理研究所/碩士(2006年)

摘要

隨著消費方式轉變為「少量多樣且一次購足」的型式，生產者也從一貫作業之生產邏輯調整為專業分工的方式。因此，物流產業在供應鏈中扮演角色變得更為重要。為達成最佳化物流系統效能、提升服務水準及設計良善配送路徑規劃排程之目的而衍生專業物流中心(Distribution Center, DC)的設置需求，也成為必然之發展趨勢。物流中心在商品的實體配銷過程中扮演集中分配的角色機能，內部作業中，尤其以訂單處理與揀貨作業最為重要與繁雜，訂單揀貨能否於有限時間內完成，密切關係物流中心之運作效率與服務水準。但供應鏈中存在著需求不確定性，需求不確定性反應在物流中心上將產生「緊急插單(Rush Orders)」，緊急插單表示物流中心無法掌握訂單出現之時機與規則，其可能導致物流中心之庫存無法滿足該訂單之需求，使服務水準下降等問題；此外，緊急插單亦可能無法與其它訂單共同批次揀貨處理，而造成整體揀貨效率降低，故物流中心能否掌握訂單需求規則是訂單揀貨作業能否有效率進行的重要前提之一。此外，部分物流中心甚至規模擴大直接與終端顧客接觸，並延伸涉及產品設計與開發自有品牌，因此，物流中心更有主動瞭解市場需求之必要性。本研究藉由資料探勘方法中之關聯法則(Association Rule)、關聯式分類(Associative Classification)以及序列樣式探勘(Sequential Pattern)方法，建構物流中心訂單需求模式。協助物流中心瞭解訂購產品之時機、訂購產品之訂單與顧客的屬性以及產品訂購序列特徵，期望藉由掌握訂單需求與減少緊急插單，以降低訂單別揀貨頻次與訂單需求不確定性之影響，進而達成提升訂單批次揀貨效能及主動瞭解顧客需求等之目的。

關鍵字

物流中心 ； 資料探勘 ； 關聯法則 ； 關聯式分類 ； 序列樣式 ； 時間區間序列樣式

並列摘要

With the way of customers’ purchasing behavior being converted to one stop shopping, small size and high diversity, manufacturers and distributors are changing their operations from vertical integration to specialized divisions. Thus, the role of logistics is becoming more important in the supply chain. To optimize the efficiency of a logistics system, to raise service level, and to design a good routing plan, the distribution center (DC) is being desired sharply, the need of which is increasing and the setting of which is a virtually inevitable trend. DC plays a role of centralized distribution in the process of physical products’ distribution. Among internal operations in DC, order processing and order picking are especially the most important and complicated. Whether the order picking is finished within the limited time highly influences the operation efficiency and service levels. Due to demand uncertainty existed in supply chain, such uncertainty reflected in DC is “rush order.” Rush orders represent that DC cannot know the timing and the rule of order placing well, which may result in that the inventory in DC cannot fulfill the demand of rush orders, and thus decrease the service level. Additionally, rush orders cannot be batch picked together with the other orders, which decreases the efficiency of order picking. Therefore, provided that DC can obtain the rules of order demand, the operations of order picking can be conducted more efficiently. Moreover, the scale of some DCs is so huge that such DCs can contact with end consumers, design their own products, and develop their own Ordnance Bench Mark (OBM). Consequently, it is imperative for DC to understand the market demand actively. In this thesis, data mining techniques including association rule, associative classification, and sequential pattern are applied to establish an order demand model for DC, which can aid DC to realize the timing of product purchasing, and the properties of customers and orders of purchased products, and the sequential characteristics of purchased products. The proposed approach in this thesis intends to acquire the order demand and decrease rush orders, which can decrease the frequency of single-order picking and the effect caused by order demand uncertainty, and then increase the frequency of batch picking and understand the customers’ needs actively.

並列關鍵字

Distribution Center ； Data Mining ； Association Rule ； Associative Classification ； Sequential Pattern ； Time-Interval Sequential Pattern.

參考文獻

[1] Agrawal, R., Imielinski, T., and Swami, A., 1993a, Mining Association Rules between Sets of Items in Large Databases, ACM SIGMOD Conference Washington DC, USA, 254-259.

[3] Agrawal, R., and Srikant, R., 1994, Fast algorithm for mining association rules, The International Conference on Very Large Data Bases, 487-499.

[7] Chen, X., and Ilias, P., 2000, Discovering temporal association rules: algorithms, language and system, Proceedings of the 16th International Conference on Data Engineering, 2000, 306-306.

[8] Cho, Y. H., Kim, J. K., and Kim, S. H., 2002, A personalized recommender system based on web usage mining and decision tree induction, Expert Systems with Applications, 23(3), 329–342.

[9] Chen, Y. L., Chen, S. S., and Hsu, P. Y., 2002, Mining hybrid sequential patterns and sequential rules, Information Systems, 27(5), 345-362.