中文摘要：

聚类是按照事物的某些属性，把其聚集成类，使各类问的相似性尽量小，类内相似性尽量大。现在使用的一些聚类算法大多效率不高、聚类速度慢。文中在改进LBG算法的基础上提出了一种新的聚类算法，克服了传统的LBG算法的缺点，具有准确性高、测试时间短的优点。现将它应用于股票数据的预测分析中，实验结果表明这种新的聚类算法，相较于其它聚类算法能够取得更好的结果。

英文摘要：

Clustering assembles things according to some of their attribute, minimize the comparability among clusters and maximize the comparability inside each cluster. Many algorithms now in used have some defects such as inefficient and slow. This text forwards a new arithmetic based on the improvement of LBG arithmetic. This new algorithm has gotten over the shortcomings of the traditional LBG with high veracity and needs short test time. Now it applied on stock data forecast analysis. The experiment shows this new clustering algorithm will yield better outcome than the old ones.