Abstract: The current R-tree spatial clustering algorithms use a predefined k-value for clustering, and choose the initial clustering centers arbitrarily. The clustering results are easily dominated by the initial k-value and the outlier data. In order to solve these problems, this paper proposes a novel spatial clustering algorithm called DKSC (dynamical k-value spatial clustering algorithm), which dynamically determines the optimized k-value when constructing the tree. By choosing an optimized k-value, the algorithm allows the spatial data in the same subspace to be organized into the same sub-tree, and builds an efficient R-tree layer by layer from the root to leaves. The experiments conducted with real and simulated datasets show that the proposed algorithm can build optimized R-tree spatial indexes and improve retrieval efficiency.

Key words: spatial data, R-tree, spatial index, clustering algorithm

摘要： 目前采用的R-树空间聚类技术使用指定k值的聚类算法，初始聚类中心随机或指定选取。这样聚类的结果受初始k值影响，且易受离群空间数据的干扰。为解决上述问题，根据空间数据分布的特点，提出了动态确定k值的空间聚类算法(dynamical k-value spatial clustering algorithm, DKSC)。该算法通过聚类划分空间数据，把同一子空间的数据组织在同一个子树下，从根节点到叶子节点逐层构建R-树，形成高效的R-树空间索引。分别用真实和模拟的空间数据集进行了实验，结果表明该算法优化了构建的R-树空间索引，且具有更高效的查找效率。

关键词: 空间数据, R-树, 空间索引, 聚类算法