基于模型的不等间隔时间序列聚类算法研究

Model-based clustering algorithm for time-series with irregular interval

现有的聚类算法一般只能处理以固定间隔表示的数据类型,而忽略了时间轴的变化。基于倒谱距离测度和自回归条件持续期(ACD)模型的聚类方法综合了计量模型的参数估计和聚类的非参无监督分类的优点,是一种适合处理不等间隔时间序列的技术。实验结果证明这种方法是有效的,从中得出的结论与市场微观结构理论也是相吻合的。

Most existing clustering methods can only work with fixed-interval representations of data,ignoring the variance of time axis.A model-based clustering approach using cepstrum distance metrics and Autoregressive Conditional Duration （ACD） model is proposed, it integrates the merits of parametric econometrics and non-parametric clustering,and is fit for time series with irregular interval.Experimental results show that this method is generally effective in clustering irregular space time series,and conclusion inferred from experiment results is agree with the market microstructure theories.