一种融合多算法的面状要素空间聚类方法  

A spatial clustering method fusing multiple algorithms for area feature

作　　者：杨杰[1,2] 张新长[2] 何显锦[2] 

机构地区：[1]吉首大学生态旅游湖南省重点实验室∥城乡资源与规划学院,湖南张家界427000 [2]中山大学地理科学与规划学院,广东广州510275

出　　处：《中山大学学报（自然科学版）》2017年第3期134-139,共6页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：国家自然科学基金重点项目(41431178);湖南省社科基金(14YBX026);生态旅游湖南省重点实验室开放基金(JDSTLY201207)

摘　　要：空间聚类各算法均有各自的优缺点,可通过融合各算法优点达到对空间聚类算法改进优化的效果。提出了一种融合多算法的面状要素空间聚类方法。该方法利用遗传算法等优化算法优化K-means算法的初始聚类中心,利用基于密度的快速聚类算法选取K-means算法的k值,最终利用改进的K-means算法得到空间聚类结果。此外该方法针对遗传算法易受初始种群影响、运算效率低等缺陷进行了改进。经实验验证,文中方法结果稳定,算法效率、结果精准度较传统算法提升明显。Each spatial clustering algorithm has its own advantages and disadvantages. Spatial clustering algorithms can be improved and optimized through the fusion of the algorithms＇ advantages. A spatial clustering method fusing multiple algorithms for area feature is proposed in this paper. This method opti- mizes the initial cluster centers of K - means algorithm by using genetic algorithm and other optimization algorithms, selects the k value of K -means algorithm by using a fast clustering algorithm density - based, and then obtains spatial clustering results with improved K -means algorithm. It improves the genetic algorithm which is easy to be affected by the initial population and has low efficiency. The experimental results indicate that the method is steady, and is more efficient and accurate compared to the traditional algorithm.

关 键 词：空间聚类 融合 智能优化算法 存储池 

分 类 号：P208[天文地球—地图制图学与地理信息工程]