摘要
为了提高K均值聚类算法的质量与收敛速度,提出一种基于小波变换的栅格数据聚类新算法。该算法利用小波分析塔式算法得到的顶层栅格数据,既较好地保留原始数据的特征信息,又大幅减小了数据量,在保证聚类质量前提下,提高了算法的收敛速度;针对分解后的低频数据应用K均值算法,得到后续迭代所需的聚类中心初值,避免了初值选择的盲目性。试验表明,该算法具有计算效率高、稳定性好、聚类质量有保证等优点。
In order to promote the quality and efficiency of K-means clustering algorithm, a new wavelet-based clustering algo- rithm has been proposed in this paper. By utilizing the advantage of saving the key feature information after decomposing of raster data with wavelet transformation, the initial values of clustering center have been gotten by applying K-means algorithm to decomposing low frequency raster data. The experiment shows that the algorithm has advantages of high efficiency, better stability and higher clustering quality.
出处
《地理与地理信息科学》
CSCD
北大核心
2008年第4期36-38,56,共4页
Geography and Geo-Information Science
关键词
空间聚类
小波变换
栅格数据
spatial clustering
wavelet transformation
raster data
