臺灣地區國中教師數預測模式

Abstract

近年來，教師供需失調的問題時常發生，除了不可預測的因素無法預料外，預測 工具的不穩健更是重要原因之一。本研究以單變量時間數列、狀態空間模式及神經網路來預 測民國 77 ∼ 82 學年度國中教師數，並對以上三種模式之預測效率做一比較。我們發現， 單變量時間數列預測誤差較大，但以狀態空間模式及神經網路模式來預測，預測誤差小很多 。因此，狀態空間模式及神經網路模式的精確預測法可以推廣，以供師資培育計劃參考。

Recently, the problem of the demand for and supply of teachers in junior high schools has been paid more attention in education administration. An accurate forecast of the number of teachers needed in junior high schools may heavily affect educational policy. In this paper, we use the univariate time series analysis, state space and Neural Networks to forecast the number of teachers in the junior high school of Taiwan Area during a period from 1988 to 1993. It was found that the state space and Neural Networks exhibit a much more successful forecast than the univariate ARIMA model.