數位相機業關鍵維修備品之需求預測與庫存策略研析

Abstract

現今企業面對的商業環境快速而多變，為滿足顧客需求，預測是回應 市場動態變化及有效降低存貨成本的重要手段，不過適合的預測需求方式 卻因產品的差異性而有顯著地不同。特別是如何有效管理產品生命週期短 暫的備品零件存貨更是仰賴精確的預測和動態存貨調整機制。有鑑於此， 本研究目標在針對數位相機關鍵備品零件-印刷電路板(簡稱機板)，藉由整 合需求預測與動態存貨調整機制的模型提出有效的庫存管理策略。 根據數位相機維修中心的實際維修紀錄(機板的需求數量)，藉由過去 銷售的數位相機銷售數量分別以週、月、季三種不同的規劃範圍之迴歸模 型回推機板需求量作預估，並以 MAPE（絕對平均誤差）做模型校估與比較。 再者，結合需求預測結果的動態存貨調整模型，以演化式計算為基礎進一 步發展最佳相關參數以致於最小化總存貨成本，包括持有成本、訂購成本、 缺貨成本及運輸成本。 整合存貨模式結果顯示，以月作資料處理範圍的預測技術的績效表現 最佳，可有效降低存貨成本表明該模型的適用性。

The current commercial environment is rapidly changed and diversified.In order to meet the request of customers, forecasting is important in the response to dynamic changes of market demand and to effectively curtail the inventory cost. However, it is well recognized that demand forecasting for different products is dramatically different. Especially, the successful inventory management of the spare parts of short-lived digital products extremely rely on the accurate forecast and dynamic inventory adjustment mechanism. Based on this, this study aims to propose effective inventory management strategies for the critical spare parts of digital cameras, Printed Circuit Board (PCB), based on an integrated demand forecasting model and dynamic stock adjustment mechanism.

According to the real maintenance records (quantity demanded of PCB) of a digital camera maintenance center, three regression models under various planning horizons: weekly, monthly, and seasonal, are respectively estimated by regressing quantity demanded of PCB on the sales quantity of digital cameras in previous periods. These models are then further compared in terms of MAPE (mean absolute percentage error). Furthermore, a dynamic stock adjustment model based on the forecasted demand is then developed along with the related parameters optimally tuned by evolutionary computation so as to minimize the total inventory cost, including the holding cost, ordering cost, shortage cost, and transportation cost.

The results show that the integrated inventory model based on the monthly forecasting technique performs best, which can effectively curtail inventory cost, suggesting the applicability of the proposed model.