企業銷售預測之方法與實証研究――以台電公司為例

徐守德 ； 李鎮旗

《管理評論》 13卷1期 (1994/01) Pp. 23-56

https://doi.org/10.6656/MR.1994.13.1.CHI.23

摘要

本研究在於探討單變量ARIMA、轉移函數、單純預測等預測方法，以了解企業銷售性質及發展最適銷售預測模式，以台電公司為例，作實證研究。
研究結果發現：1).單變量ARIMA與轉移函數相比較，其預測準確度相差有限，這顯示若預測本身只是欲推斷未來各月之售電量，而不探求其變化原因時，則可以單變量ARIMA做預測即可；2).各轉移函數所求得之預測準確度相差不大，表示轉移函數預測法，有一套完整的邏輯推論基礎，所以適用於各種銷售預測情境；3).各種單純預測法間準確度相差較大，表示在建立銷售單純預測模式時，宜多衡量銷售資料特性方能得到良好的預測模式。

並列摘要

This article mainly discusses univariate ARIMA, transfer function, and naive methods for forecasting enterprise sale and takes Taiwan Power Corporation as an example.
It is observed that 1). In comparing with transfer function, univariate ARIMA is quite accurate in forecasting; 2). The forecast accuracy among transfer functions does not have too much difference, which indicates that forecasting via transfer function is logically deductive and can be applied for sale forecasting; 3). The forecasting differences among naive methods are substantial, which reveals that the user should have good knowledge of the series data.