符号化近似SAX在时序数据挖掘中的应用研究

聚类是数据挖掘研究中最常见的一种方法,可以作为规则发现、异常发现等其它数据挖掘操作的基础,一直以来都是数据挖掘的研究热点之一。股票数据是一种典型的时间序列数据,利用股票数据进行时间序列数据挖掘的研究既有一定的实际应用价值,也是国内外的热点问题之一。文章首次将一种新型符号化方法SAX1]应用到标准普尔500指数的股票数据的聚类研究中,使用传统的欧氏距离和动态时间弯曲两种时间序列相似性度量方法进行实验。实验结果表明将SAX应用到股票数据聚类操作,可以得到更好的趋势聚类效果和更高的效率。

Application Research of a New Symbolic Approximation Method-SAX in Time Series Mining

Clustering is one of the most common data mining methods,being a hot topic in its own right as an exploratory tool,and also a subroutine in more complex algorithms such as rule discovery and abnormal discovery.As a typical time series data,stock data has been widely used in data mining research.In this paper,a new symbolic method-SAX1] is used in stock data clustering analysis.We apply the new method on stock data which is obtained from the Standard & Poor 500 index.Moreover,we use two similarity measure method including Euclidean and Dynamic Time Warping in our experiments.The experiments result shows that clustering can better focus on the whole trend and the efficiency can be improved with the help of SAX.