• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2005, Vol. 27 ›› Issue (10): 81-84.

• 论文 • 上一篇    下一篇

一种基于回归分析的空间划分方法

滕明贵[1,2] 马献章[2] 曾建勤[1]   

  • 出版日期:2005-10-01 发布日期:2010-06-24

  • Online:2005-10-01 Published:2010-06-24

摘要:

针对子空间连通性约束情况下空间对象的局部回归分析问题,本文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。

关键词: 回归分析 空间划分 边界调整

Abstract:

To solve local regression with the connectivity constraint of spatial objects in sub-spaces, this paper proposes an algorithm for clustering spatial o bjects based on the criteria of the minimum square error sum. After initializing the original partitioning, the boundaries of sub-spaces are adjusted to  construct new partitioning. Such process will continue until the criteria function converges.

Key words: (regression analysis;spatial partitioning, boundary adjustment)