運用貝氏管制圖於定期訂購存貨系統之研究

Abstract

資訊科技的進步促使資訊的傳遞變得更加快速與便利，同時也改變消費型市場的樣貌。消費者可以輕易地取得資訊與擁有更多的消費選擇，此也使市場競爭愈趨激烈。而企業為了迎合消費者多元的需求，必須不斷提高供應鏈之效率來滿足客戶，過去有許多文獻在探討如何設計一套模型來預測消費者之需求，其中，貝氏方法提供了不同於傳統點估計之預測模式，也有文獻指出，管理者可透過學習以及過去的經驗來更新先驗資訊，以提高需求預測之準確型；然而，有預測必定有誤差，故以存貨管理的角度來說，建立管理監控機制實屬重要，然而過去之研究多半以頻率學派管制圖結合存貨管理為主，吳致賢(2017)首次提出結合貝氏管制圖與追蹤訊號於存貨管理之需求監控，然而該研究假設無訂購時間之下來評估貝氏管制圖於存貨管理之應用，但一般來說，在探討存貨管理之議題，訂購前置作業時間這項變數扮演很重要的角色，並將之視為常數或隨機變數；且實務上，企業認為訂購前置時間在存貨管理的範疇上為一重要議題，因此本研究欲以其研究為基礎，探討若考慮訂購前置時間之需求於貝氏的先驗資訊中的影響，並進一步比較貝氏管制圖與傳統定期盤存制度的差異。 本研究以某食品公司的方便麵銷售為例，針對兩種不同需求變異之口味進行分析，同時也比較訂購前置時間為一與零的情境。分析數據結果後發現，需求變異對於兩系統之影響較小，且貝氏管制圖相較於傳統管制圖會產生較低的存貨水準，但相對地會產生較嚴重的缺貨現象；在無訂購前置時間之下，貝氏管制圖的總成本比傳統管制圖來的高，但當訂購前置時間為一期時，貝氏管制圖可擁有較傳統管制圖較低的總成本。本研究認為，多考量前置時間的需求時，共擔效果(Pooling effect)發揮作用，藉由兩期需求合併來減緩因變異所帶來的缺貨衝擊，故使在擁有相同服務水準之下，貝氏管制圖在總成本的績效表現優於傳統定期盤存制系統。

With the technology advances, the information delivery is becoming faster and more convenient. Meanwhile, it also changes the consumption market, and the customers can get the information easier and have a variety of choices in this era. The market competition, therefore, is being more intense than before. In order to cater to the diverse needs of customers, companies have to improve the efficiency of the supply chain to satisfy the customers’ needs constantly. In the past, there were a lot of literatures on how to design a model to predict the market demand. Among them, Bayesian method provides a prediction model different from traditional statistical method point estimation. It is pointed out in the literature that managers can update prior information through their learning and experiences, which can help improve the accuracy of demand forecast. But we realize that there must be errors in forecasting, in terms of inventory management, it’s important to establish the management monitoring mechanism for the inventory management. Most of the previous studies have focused on the frequency school control chart combined with inventory management. Wu (2017) first proposed the monitoring process for inventory with Bayesian control chart combined with tracking signal. However, in most of literature dealing with inventory problems, they considered the lead time in their model which they viewed as constant or stochastic variable. Moreover, in practice, ordering lead-time is an important issue in the company. Hence, based on Wu (2017) research, we want to understand the impact of considering the demand within lead-time in the prior information and further compare the difference between Bayesian control chart and the traditional periodic inventory system. This study takes the sales data of instant noodles as an example to analyze two tastes of different demand varieties, and compares the situation in which lead-time is zero and one. After analyzing the data, it is found that the variation of demand has less impact on the two inventory monitoring system but Bayesian control chart will produce lower inventory level; in the contrary, it will cause more stock-out level. Without considering lead-time, the total cost of Bayesian control chart is higher than traditional chart, but lead-time is one, the total cost of Bayesian control chart is lower. This study thinks that “pooling effect” plays a role in considering lead-time situation. Under the same service level, it can combine two periods of demand to mitigate the stock-out shock caused by demand variation. Therefore, in terms of total cost, the Bayesian control chart is superior to the traditional periodic inventory system when considering lead-time in the model.