Abstract: In this paper a segmentation model and algorithm based on R/S analysis is proposed for the application of rescaled range analysis in time series data mining.The series can be divided into some subsequences with proposed algorithm according to the character of clustering of volatility and self-similarity.Experimental results show that the volatility change tendency of time series can be obtained by the method.

Key words: time series, Rescaled Range（R/S） analysis, clustering of volatility, self-similarity

摘要： 针对重标级差分析法（Rescaled Range Analysis，R/S）在时间序列挖掘中的应用，提出了一种基于R/S分析的时间序列分割模型和算法，该算法能够根据序列波动的聚集性和自相似的特征，将序列分割为多个子序列。实验结论表明该方法可以发现时间序列的波动变化规律，方法有效、正确。

关键词: 时间序列, 重标级差分析, 波动聚集, 自相似