时间序列依存聚类及其在金融领域中的应用研究.pdfVIP

ISSN 计算机研究与发展 1000—1239／CN11-1777／TP of Researchand JournalComputer Development 42(增刊)：112～116，2005 时间序列依存聚类及其在金融领域中的应用 何洋波 梁 循 (北京大学计算机科学技术研究所北京100871) (heyangbo@icst．pku．edu．cn) TimeSeriesBasedon DistanceandIts in Clustering pendence Applications FinancialMarket He and Xun YangboLiang (Instituteof Scienceand 100871) Computer Technology，PekingUniversity，Beijing on with Abstract basedcorrelationdistance time correlation Clustering assignsseri．es highpositive coefficientsintothesame andthosewith correlationcoefficientsintodifferent negative group high groups． In and correlationsthe betweentimeseries．A fact，bothpositivenegative implydependencerelationships new distance，is to the oftime similaritymeasure，dependenceproposedrepresentdependencerelationships series．An of theinitialcenter of is to k—meanscluster．The algorithmchoosing pointsgroupsgivenperform thetimeseriesthat influenceeach thoseinfluenced proceduregroupstogether strongly other，and proposed to acommontimeseries．Inorder those features measurebasedon stronglyby interpretwhy group exist，a maximum estimationis toevaluatethe causes．The withrealworld posterior provided possible experiment 17stocksfromthe stock thatthe isa causethat data，6 Shanghaiexchangemarket，s