對於超級市場的採購人員而言,如何決定最佳的商品訂購量是一個非常頭痛的問題,因此需要發展適當的預測模式來推估市場的需求量。本研究針對超市販售的易腐性商品,分別採用時間序列預測方法、灰色理論和類神經網路,以及多重指標模式建構合乎其商品特性的需求預測模式,並且以S超市2005年至2006年七月份某品牌鮮乳的銷售資料為樣本進行模擬分析。結果顯示多重指標模式預測的準確度和時間序列方法相差無幾,且明顯優於灰預測GM(1,1)模型和類神經BPN網路模式的預測結果,特別是採用週期加權模式具有較佳的效果,利用該預測模式來決定易腐性商品每日的供貨量,應可兼顧服務水平且降低存貨數量,為企業創造更高的利潤。
How to determine the optimal order quantity of goods is a real headache for the procurement staffs in supermarkets, therefore they need to develop an appropriate forecasting model to estimate the real demands on the market. This study is carried out in S supermarket and focuses on perishable goods. We use time-series forecasting method, grey theory, neural network, and weighted multi-indicator model to construct demand forecasting models with the characteristics of perishables, and select milk sales from 2005/1/1 to 2006/7/31 as samples for analysis. The results showed that the forecast effect from weighted multi-indicator model is almost the same as that from time-series method, it is also superior to those from the GM (1,1) model and the BPN forecasting model. The forecast effect is even better when using the weighted weekly indicator. These supermarkets' procurement staffs utilizing this forecasting model to determine the daily supply of perishable goods should be able to achieve higher service level and lower number of inventories, and thus create more profits for the enterprise.