Abstract: According to the neglect problem of the importance of attribute data while spatial data clustering and the tendency exploration of spatial feature data, a trend function is established for describing the attribute value change with spatial location. Then, second-order model is constructed by referencing to the variation function. In view of the above, a similarity function integrated spatial distance and attribute difference is built. A treatment scheme regarding of angle tolerance is discussed under stationary hypothesis. Finally, a cluster model named K-Trend is set up with taking trend function as a core. The results show that the K-Trend cluster method has high quality, is seldom affected by sample size, and has moderate time-consuming. All of these features improve the practicability of spatial data cluster.

Key words: trend function, spatial data, cluster

摘要： 针对空间数据聚类不重视属性数据利用的问题，在探讨空间属性数据趋向性的基础上，构建刻画属性值随空间位置变化的趋势函数，形成了借鉴于变异函数的二阶模型，并据此建立一个集成空间距离和属性差异的相似度函数，探讨了满足平稳假设和角度容限处理方案，形成了一个以趋势函数为核心的聚类模型K-Trend。实验表明：该算法聚类结果质量高，不受样本影响，且耗时适中，进一步提高了空间数据聚类的实用性。

关键词: 趋势函数, 空间数据, 聚类