使用列聯表建立需求管理為基之離散選擇統計模型

Abstract

需求預測是營運管理中重要的課題，透過預測下游市場需求對產品價格的反應，得以建立良好的需求管理。在許多產業，市場需求會受到個別廠商的牌價制定決策以及當時經濟情況所影響，廠商可以透過對牌價的設定來影響供應鏈下游廠商的採購決策。傳統的需求預測模型較多偏重於產經數據對於銷量的影響,較少深入探討廠商如何影響市場預期。廠商若能觀察市場的經濟情形並制定適當的牌價，就可以達到有效的需求管理。本文以離散選擇統計模型為基礎，根據過去接單量之狀態遷移路徑，搭配層級貝氏計量模型，建立異質性轉移機率矩陣，進而預測接單量狀態的轉移，藉此優化需求管理。

Demand forecasting plays an important role in operations and supply chains. Firms can establish a full-fledged demand management if they can forecast the demand from downstream manufacturers responding to the price of the goods accurately. In many industries, firm’s policy of price tags and current economy situations have a great influence on market demand, and they can affect the purchasing decisions from the downstream manufacturers by declaring suitable price tags. Conventional empirical research on demand forecasting relies on investigating the effects of economic data on the order quantity rather than exploring how firms can manipulate the market expectation. Firms can achieve effective demand management by declaring suitable price tags after examining the economic situation. This study is based on discrete choice model, and construct heterogeneous transition probability matrix using the historical sales data of state migration. With our method, firms can adopt suitable strategy for price tag to forecast the state migration of order quantity accordingly.