文献类型：期刊文献

英文题名：A Clustering algorithm of high-dimensional data based on sequential psim matrix and differential truncation

作者：Wang, Gongming[1]; Li, Wenfa[2]; Xu, Weizhi[3]

第一作者：Wang, Gongming

通讯作者：Wang, Gongming

机构：[1] Institute of Biophysics, Chinese Academy of Sciences, No. 15 Datun Road, Beijing, China; [2] College of Information Technology, Beijing Union University, No. 97 Beisihuan East Road, Beijing, China; [3] School of Information Science and Engineering, Shandong Normal University, No. 88 East Wenhua Road, Jinan, China

第一机构：Institute of Biophysics, Chinese Academy of Sciences, No. 15 Datun Road, Beijing, China

年份：2018

卷号：11335 LNCS

起止页码：297-307

外文期刊名：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

收录：EI(收录号：20185106273723)

基金：Acknowledgments. This work is partly supported by the National Nature Science Foundation of China (No. 61502475, 61602285) and the Importation and Development of High-Caliber Talents Project of the Beijing Municipal Institutions (No. CIT & TCD201504039).

语种：英文

外文关键词：Trees (mathematics) - Heuristic algorithms - Matrix algebra

摘要：For high-dimensional data, the failure in distance calculation and the inefficient index tree that are respectively derived from equidistance and redundant attribute, have affected the performance of clustering algorithm seriously. To solve these problems, this paper introduces a clustering algorithm of high-dimensional data based on sequential Psim matrix and differential truncation. Firstly, the similarity of high-dimensional data is calculated with Psim function, which avoids the equidistance. Secondly, the data is organized with sequential Psim matrix, which improves the indexing performance. Thirdly, the initial clusters are produced with differential truncation. Finally, the K-Medoids algorithm is used to refine cluster. This algorithm was compared with K-Medoids and spectral clustering algorithms in two types of datasets. The experiment result indicates that our proposed algorithm reaches high value of Macro-F1 and Micro-F1 at the small number of iterations. ? Springer Nature Switzerland AG 2018.