Abstract: According to the characteristics of Chinese short message text, this paper presents a clustering process based on algorithms of the optimal number, which can be obtained multiple statistical information by scaning the data only once, and using increment in-depth profile analysis so as to obtain corresponding total number of class cluster, obtained optimal solution. Experimental result shows that the method has advantage with highly-quality than other methods.

Key words: clustering, cluster number, increment, division

摘要： 针对海量中文短信文本的聚类簇数的确定问题，提出一种基于聚类过程的短信文本最佳聚类数确定方法。通过扫描一遍数据即可获得多个统计信息，利用增量逐层划分得到最优划分所对应的簇类数，求出最优解。实验结果表明，与其他方法相比，该方法的分类效率较高。

关键词: 聚类, 簇数, 增量, 划分

CLC Number: 

• TP311