基于形态拟合的时间序列距离计算

针对现有的距离度量方法度量准确度低且计算效率低的问题,提出了基于形态拟合的距离度量算法.该算法使用滑动聚集平均近似方法对序列进行分段降维处理,计算降维后的分段序列的动态弯曲路径,并计算处于动态弯曲路径上的分段序列之间的欧式距离,以所有分段序列的欧式距离的累积值作为最终的距离计算结果.实验表明基于形态拟合的距离度量算法具有度量准确度高且计算效率高的优点.

Calculating the distance of time series by form-fitting

Existing distance measures were characterized by low accuracy and efficiency.Thus,a form-fitting distance measure was proposed which combined the advantages of both Euclidean distance and dynamic warping distance.Sliding gathered average approximation method on the sequence of piecewise dimensionality reduction treatment was used in the algorithm.The dynamic warping path between the dimensionality reduction sequences was calculated.The piecewise Euclidean distance between the sequences in the dynamic warping path was calculated.The value of the Euclidean distance of all segments sequence was cumulated as the final results of the distance calculation.It has the advantages of high accuracy and efficiency.