基于形态距离及自适应权重的相似性度量

Similarity measure based on morphological distance and adaptive weights

针对传统的动态时间弯曲算法的性能容易受到离群点以及局部噪声点的影响,同时对于复杂数据的处理能力较差,提出基于形态距离及自适应权重的相似性度量算法。该算法首先利用l1趋势滤波对原始待比较序列进行降维、压缩;其次引入形态距离计算两时间序列的距离矩阵;最后利用自适应赋权的距离函数抽取出各个子序列所含的信息量差异并结合动态时间弯曲完成最终时间序列相似度量。实验表明,该算法有更强的鲁棒性,能够更好地利用序列的形态特征完成宏观的相似性度量,同时在处理复杂数据时更加精确、高效、稳定。

The performance of the traditional dynamic time bending algorithm is susceptible to outliers and local noise points,and the processing capacity of complex data is poor.In this regard,this paper proposed a similarity measure based on morphological distance and adaptive weight.The algorithm first used the l 1 trend filter to reduce the dimension and compression of the original comparison sequence.Secondly,the algorithm introduced morphological distance to calculate the distance matrix of two time series.Finally,the algorithm used the distance function of adaptive weight to extract the difference of information contained in each sub-sequence and completed the final time series similarity measure with dynamic time bending.Experiments show that the algorithm has stronger robustness and can make better use of the morphological features of the sequence to complete the macro similarity measure,while dealing with complex data more accurate,efficient and stable.