面向动态时间弯曲距离的灰关联聚类方法研究  

Grey incidence clustering method based on dynamic time warping distance

作　　者：张起荣 刘歆[2] 闫一[2] 

机构地区：[1]贵州工程应用技术学院信息工程学院,贵州毕节551700 [2]重庆邮电大学软件学院,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2015年第4期563-568,共6页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金(61309014);重庆市基础与前沿研究计划项目(cstc2013jcyj A40009);重庆市教委科学技术研究项目(KJ1400412);毕节学院高层次人才科学研究项目(G2013006)~~

摘　　要：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 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 une- qual-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 doesnt need manual intervention and possesses stronger ro- bustness. Finally, the experiment shows that the clustering result of this novel method is more efficient when handling in- consistent-length data sequences.

关 键 词：灰关联聚类 灰关联分析 动态时间弯曲距离 灰关联度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]