【摘要】 在传统的水文时序数据研究中,我们通常只关注单个测点的时序数据,这不仅造成数据大量的冗余,还大大增加了工作的繁琐度。本文针对时间序列数据聚类的统计特征和结构特征,基于滑动窗口特征提取算法提出了改进的K-means聚类方法,来探求水文时间序列数据是否在空间上存在某种关联,并在此基础上对多水文站数据进行关联异常检测。更多还原

【Abstract】 In the study of traditional hydrological time-series data, it usually only focuses on a single point of time-series data. This not only causes a large number of redundant data, but also greatly increases the complicated degree of work. In this paper, according to the statistical characteristics and structure features of time-series data clustering, K-means clustering method which is based on feature extraction algorithm of sliding window is put forward to explore whether there is a correlation between hydrological time series data in the space, and anomaly detect the multiple hydrologic data on the basis of it.更多还原