基于数据取样的DBSCAN算法

取类是数据挖掘领域中的一个重要研究课题。聚类技术在许多领域有着广泛的应用，基于密度的聚类算法DBSCAN是一种有效的空间聚类算法，它能够发现任意形状的类并且有效地处理噪声，用户只需输入一个参数就可以进行聚类分析，但是，DBSCAN算法在对大规模空间数据库进行聚类分析时需要较大的内存支持和I/O消耗。本文在分析DBSCAN算法不足的基础上，提出一种基于数据取样的DBSCAN算法，使之能够有效地处理大规模空间数据库。二维空间数据测试结果表明本文算法是可行、有效的。

SDBSCAN :A SAMPLING-BASED DBSCAN ALGORITHM FOR LARGE-SCALE SPATIAL DATABASES

Clustering, a useful technique for data mining, has promising applicati on perspective in many fields, such as pattern recognition, image processing and data compression etc. DBSCAN is a density-based clustering algorithm that can d iscover clusters of arbitrary shape and effectively handle noise, and requires o nly one input parameter. However, while dealing with large-scale databases, DBS C AN requires a lot of memory and can incur high I/O overhead. In this paper, we p resent a sampling-based DBSCAN algorithm(SDBSCAN)for clustering large-scale da tabases. Experimental results demonstrate that our algorithm is feasible and eff ective.