面向动态时间弯曲距离的灰关联聚类方法研究

灰色关联聚类是灰色系统分析的重要研究领域.现有的灰关联聚类方法在处理不同长度序列数据时,通常采用补齐或删除数据方式进行处理,增加了系统的不确定性.在动态时间弯曲距离基础上,提出了一种新的灰关联聚类方法,通过计算序列间距离矩阵的最短路径来度量相似程度,最终完成序列的灰关联聚类.该方法无需对序列数据进行人工干预,具有更强的鲁棒性.实验结果表明,在处理不同长度数据序列时,聚类结果更为准确.

Grey incidence clustering method based on dynamic time warping distance

Grey incidence clustering method cannt deal with data sequences with different length. These methods usually pad up the shorter data sequence or delete some redundant data, and that will increase the uncertainty of grey system. To solve the problem, a novel grey incidence clustering method by introducing dynamic time warping distance used for unequal-length sequences processing is proposed. It can measure the similarity between sequences by computing the shortest path of distance matrix to achieve grey clustering. This method doesn''t need manual intervention and possesses stronger robustness. Finally, the experiment shows that the clustering result of this novel method is more efficient when handling in-consistent-length data sequences.