基于最小拟合误差平方和准则的空间聚类

在一些回归分析问题中，数据来自于空间对象的非空间属性．许多问题中不用考虑空间属性，而直接分析非空间属性．如果在整个问题空间中，对象空间差异较大，需要将空间对象划分为若干子空间，子空间对应的局域回归模型，可以减小空间差异性的影响．针对子空间连通性约束情况下空间对象的局域回归分析问题，提出基于最小拟合误差平方和准则的空间划分方法，从一个空间的初始划分开始，按照拟合误差平方和下降的原则调整子空间边界，获得新的空间划分和对应的回归模型，不断迭代直到准则函数收敛．

Clustering Spatial Data Based on Minimum Square Fitting Error Sum

In many regression analyses, data are connected with spatial objects. Most cases ignore the spatial attributes and deal only with non-spatial attributes. When diversity between spatial objects play important role, local regressions are adopted. The space is divided into several sub-spaces, each corresponding to a local regression, which decrease the effect of spatial diversity. To solve local regression with connectivity constraint of spatial objects in sub-spaces, this paper proposes an algorithm for clustering spatial objects based on the criteria of the minimum square error sum. After initializing original partitioning, the boundaries of sub-spaces are adjusted to construct new partitioning. Such process will continue until the criteria function converges.