Abstract: In order to improve the accuracy and efficiency of Clustering Large Applications based on Randomized Search(CLARANS) algorithm, this paper combines the idea of data space division which comes from grid-based algorithm Statistical Information Grid(STING), improves the CLARANS algorithm by optimizing the selection of initial node and neighbor node, optimizing the calculation of total node replaces cost. Experimental results show that, compared with the CLARANS algorithm, the improved algorithm has better accuracy and stability for the clustering results, and significantly reduce the execution time.

Key words: Clustering Large Applications based on Randomized Search(CLARANS) algorithm, Statistical Information Grid(STING) algorithm, clustering, dissimilarity degree, data space

摘要： 为提高CLARANS算法的准确性和执行效率，利用网格聚类算法对数据空间进行划分的思想，结合统计信息网格算法，对算法初始节点和邻居节点的选择及替换总代价的计算进行改进。实验结果表明，与CLARANS算法相比，改进算法聚类结果的准确性和稳定性更高，执行时间明显降低。

关键词: CLARANS算法, 统计信息网格算法, 聚类, 相异度, 数据空间

CLC Number: 

• TP301.6