近年來由於經濟環境的快速變化，使得供給和需求之間的平衡產生了許多問題，面對詭譎多變的市場，如何準確地預測未來的供需變化，一直是學術界和產業界共同面臨的挑戰。由於連鎖便利店小而美的經營方式，對於保存期限及存放空間的要求甚嚴，尤其是鮮食商品，往往受到於保存期限、運送方式及存放條件的限制。管理者為了保持商品的新鮮度而造成訂購量不足的情況層出不窮，對於主張滿足臨時性顧客需求的便利店來說無疑是一大傷害，但過多的商品也將使便利店的營運陷入危機。因此，準確的預測模式不僅有助於提升消費者對於鮮食商品的接受度及滿意度，更可以精確的控制庫存及報廢。本研究尋找可能影響上海喜士多便利店浦東南門市鮮食商品中，飯盒之銷售量的變數，針對不同的變數組合進行複迴歸分析，建構出不同的銷售預測模式來與實際的銷售量相互驗證，以複判定係數（R square）判斷不同模式對預測銷售量解釋力的高低；杜賓－瓦森法（Durbin-Watson Test, DW）檢定相鄰殘差項間的自我相關程度；平均絕對誤差百分比（Mean Absolute Percentage Error，MAPE）評估不同模式的預測效果。並藉由平均絕對偏差（Mean absolute deviation，MAD）、均方誤差（Mean Square Errors，MSE）及泰爾不等係數（Theil’s Inequality Coefficient，THEIL）三項指標進行預測誤差的評估，便能比較出不同預測模式之優劣。最後，探討本研究預測模式所得之飯盒銷售量與店長預測的飯盒銷售量（即訂購量）之差異，所得結果顯示，本研究建構的模式5預測能力不僅比其餘預測模式為佳，更以7.69％的預測飯盒銷售量之報廢率優於店長的12.52％，並符合便利店業者所期望的5～10％報廢率。 In recent years, economic environment changes rapidly, so the balance point between supply and demand is unstable and causes many problems. How to accurately estimate the trend is a challenge for academia and industry.The requirements of shelf life and storage space is very strict for convenience store's merchandise, especially fresh foods. A good forecasting system leads to improve the satisfaction of customers, reduce destruction of fresh food, lower the cost and increase sales revenue.This study is trying to find the variables that affect the sales quantity of lunch box in Shanghai CS Convenience store. By using different combinations of variables, we can constructs sales forecasting models according to multiple regression analysis. Besides, we analyze the effects of different regression equation by R square, DW and MAPE. Then using the statistical methodology includes MAD, MSE and THEIL to verify the results and choose the best model. Finally, the model will be compared with the actual sales data.
