交通流时间序列分离方法

采用聚类分析方法对交通流时间序列进行分析可以发现典型的交通流变化模式。通常 可采用欧式距离及K均值算法进行时间序列聚类,但经分析发现单凭此方法还难以实现不同变化趋 势的交通流时间序列的有效分离。针对此问题,提出了将动态时间弯曲及灰色关联度引入交通流时 间序列相似性度量,且结合层次化聚类方法对交通流时间序列进一步分离的方法。通过实验研究,发 现基于灰色关联度的层次化聚类方法能较好地实现交通流时间序列的进一步有效分离。

Traffic flow time series separation methods

By clustering of traffic flow time series, the typical traffic fluctuation patterns can be found. Generally, the euclidean distance and K-means algorithm can be used to clustering the time series, but it is hard to separate the time series with great different variability well. To solve this problem, fluctuation similarity measure, such as dynamic time warping and gray relation grade, and the hierarchical clustering algorithm were used to further separate the traffic flow time series. The experiments show that the proposed method can work and the gray relation grade measure is better suited for the problem than the dynamic time warping measure.