基于分形理论的股票时序数据离群模式挖掘研究

Outlier Pattern Mining of Stock Time Series with Fractal Theory

针对股票时间序列的特点,从离群点对股票时序数据有序性的影响角度出发,在界定分形离群点含义的基础上,利用分形理论将离群模式挖掘理解为一个优化分割问题。采用推广G-P(Grassberger-Procaccia)算法计算股票时间序列数据集的多重分形广义维数,并利用贪婪算法的思想设计了FT-Greedy算法来求解基于分形理论的时间序列离群模式挖掘优化问题的解集。实验证明,该方法能有效地解决股票时间序列离群模式挖掘问题。

According to the characteristic of stock time series, outlier pattern mining is considered as an optimization segmentation problem by using fractal theory, based on the defining fractal outlier, from the viewpoint of outlier affecting orderliness of data set of time series. G-P （Grassberger-Procaccia） algorithm is used to calculate multi-fractal and general dimension. A greedy algorithm named FT-Greedy is designed to solve the optimization problems of outlier pattern mining of time series. The experiment shows that the method is feasible to solve the problems with outlier pattern mining of stock time series.