異空間分佈點之因果迴歸分析－克利金法與AASN神經網路之比較

Regression Analysis for Different Spatial Sampling Scatter-Comparisons of Kriging Method and AASN Neural Networks

葉恰成(I-Cheng Yeh) ； 陳伯愷(Po-Kai Chen) ； 周文杰(Wen-Chieh Chou)

《農業工程學報》 53卷1期 (2007/03) Pp. 51-61

https://doi.org/10.29974/JTAE.200703.0006

摘要

本研究旨在探討當二個變數在空間下的採樣點位置不同時，如何建構其因果迴歸模型。本研究採用克利金法與AASN神經網路做為建構空間內插模型的工具。本研究採用三組個案的七個因果關係來比較各法的優劣性，研究結果顯示：(1)AASN神經網路所建立的空間內插模型遠比克利金法準確，(2)AASN神經網路所建立的異空間分佈數據的因果迴歸模型遠比克利金法準確。

關鍵字

神經網路 ； 克利金法 ； 空間內插 ； 因果迴歸 ； 空間取樣

並列摘要

The purpose of this paper is to explore, when the spatial positions of sampling points of two variables are different, how to construct their cause-and-effect regression model. This research used the kriging method and the AASN neural networks as the tools to construct spatial interpolation model. This research examined seven relations from three groups of cases to compare advantages and disadvantages of these two methods. The findings showed that (1) the spatial interpolation models constructed by AASN neural networks are much more accurate than by kriging method, (2) the cause-and-effect regression models constructed by AASN neural networks are much more accurate than by kriging method.

並列關鍵字

Neural networks ； Kriging method ； Spatial interpolation ； Cause-and-effect regression ； Spatial sampling