水文时间序列相似性查询优化算法

Similarity search optimization algorithm in hydrological time series

为了快速、准确找出给定时间段相似的水文过程,提出了一种语义相似性匹配下加权动态时间弯曲距离和标准欧式距离结合的查询优化算法。针对水文数据特点,在小波变换、特征点分段和语义符号化过程的前提下,用语义相似匹配和离散区间初步筛选候选集,使用加权动态时间弯曲距离对候选子序列进行近似匹配,利用改进欧式距离通过左右搜索法进一步优化相似结果。以鄱阳湖康山站日水位数据为例,表明了该算法在降低时间复杂度的前提下较准确地找出相似子序列。

To quickly and accurately search hydrological processes similar to a given time period, a query optimization algorithm is put forward. It combines weighted dynamic time warping and standard Euclidean distance on the premise of semantic similarity matching. According to the characteristics of hydrological data, first of all, on the premise of the wavelet transform, feature point segmentation and semantic symbol process, semantic similarity matching and discrete interval are used to filter candidate sets of time series, then similar sets of sequences are gotten from the sub-sequence of the candidate sets by weighted K-DTW approximate matching, and finally improved Euclidean distance is utilized via around search method to further optimize the result sets. Experiments on the water level of Poyang Lake Kangshan station show that the proposed algorithm is more accurate to find similar sub sequences under the premise of reducing the time complexity.