中文摘要：

提出了一种利用空间信息的分层抽样方法.首先,利用多项式回归模型估计各坐标之间的空间相关系数,然后对空间相关系数矩阵进行谱聚类分析,最终依据上述结果对总体进行分层.通过与其他抽样方法的比较证明提出的基于网络聚类的空间分层方法具有比较优良的分层效果.

英文摘要：

In this paper, we propose a stratification method utilizing the spatial information.First, we use polynomial regression method to estimate the spatial correlation between all coordinate points, then apply spectral clustering method to the matrix of spatial correlation,and finally, we can get the stratifications based on the result above. At last, we apply our method to a real data of food from the website of DianPing and confirm that our method has a good performance on stratification by comparing with other sampling methods.