R＊树结点的主元分界分裂方法  

Splitting method of principal component boundary for R＊-tree node

作　　者：李聪[1] 孙殿柱[1] 刘华东[1] 白银来 

机构地区：[1]山东理工大学机械工程学院,山东淄博255091

出　　处：《中国农机化学报》2015年第2期283-286,共4页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金项目(51075247)

摘　　要：为降低R*树结点重叠度,提高其空间利用率,通过结点特征点集方差及各子特征点集方差之和建立主元分析和结点分裂之间的联系,基于主元分析算法对特征点集进行降维处理,计算特征点集的主元向量,过特征点集中心且正交于该向量建立分界面对特征点集进行划分,将各簇数据的中心作为结点分裂的初始分裂中心,实现R*树结点分裂。实验证明,该算法具有较高的结点分裂效率,使得R*树结点重叠度降低,分裂结果较合理,显著提高了R*树构造效率和k近邻查询效率。In order to reduce the nodes overlap degree and improve its space utilization, the relationship between node splitting and principal component analysis was built based on the node feature point set variance and the sum of each feature point set variance. The dimension reduction of point set was conducted and the principal component vector was calculated. The future point set was divided using the feature point which is on the heart and orthogonal to the vector. The clustering centers were used as the initial node splitting centers to complete the node splitting process of R＊-tree. This algorithm has good node splitting efficiency which lowers the R＊-tree overlapping rate. It also improves the R＊-tree building efficiency and data query efficiency.

关 键 词：R＊树结点分裂 主元分析 主元分界面 K均值聚类 降维聚类 

分 类 号：TP391.72[自动化与计算机技术—计算机应用技术]