ANN-Krige方法及其应用

地统计学是应用数理统计学的一门新分支 ,是研究空间相关性的一种有效方法 .与传统统计学相比 ,它更注重于随机变量的空间过程 ,通过研究对象在空间上不同间隔距离的抽样点上的差异 ,定量地描述其空间变化规律 .这一理论包括两方面研究内容 :(1)半方差图 ;(2 )空间局部内插理论 .但其主要技术为半方差图模型的优化 ,这是空间局部内插理论的基础 .本文提出应用人工神经网络方法建立半方差图模型 ,进而提出空间相关性分析新方法—— ANN- Krige方法 ,并结合实例进行了应用研究 .结果表明 ,本文所提出的方法是可行的 ,其空间内插效果比传统地统计学方法明显更优 ,从而丰富和发展了地统计学的方法与理论

ANN-Krige method and its application

Geostatistics is a branch of applied mathematial statistics, and it is an effective method to study the spatial correlativity. Comparing with the traditional statistics, the theory and method of geostatistics focus on the spatial process of random variable. By studying the variance of the points separated by different spacing in the space of the subject, the spatial variance laws are described quantitatively. Geostastistics theory consists of two aspects of semi variongram and krige. But its main technique is the optimization of the semi variongram, which is the basis of spatial interpolating theory. In this paper, it was put forward how to apply the artificial neural network to establish semi variongram model, and a new analysis method of spatial correlativity called ANN Krige method was also pointed out. The results showed that the applied effect was better than traditional geostatistics theory and the new method was feasible, which enriched and developed the geostatistics theory.