ANFIS理論應用於教師人力需求預測模式之建構

Abstract

本文提出以適應性類神經模糊推論系統（Adaptive Neuro-Fuzzy Inference Systems；簡稱ANFIS）建構教師人力需求的預測模式；由於ANFIS具有高效率的學習模式，與本身分散式的網路架構，特別是對於非線性模型之建立或時間數列預測模式的建構，顯得非常的有效。經由實際資料的模擬與實測，並與新近被提出的灰色預測模式相較，發覺ANFIS即使在有限的資料筆數下，不僅具有學習速度快而且預測準確度較高，其相對平均誤差值更小的優點。

In this paper, we apply ANFIS to the modeling of a forecasting system for the demand of teacher human resources. Because of its inherent distributive architecture and the efficient learning algorithms for adapting parameters, ANFIS has been shown to be very useful in nonlinear system modeling and in time series applications. After bring tested on a set of practical data, it reveals that the learning speed of the ANFIS model is faster than that of the gray model. In addition, the average prediction error of the ANFIS model, when compared with the gray model, is greatly reduced.