两步聚类BIRCH算法.pptx

下载文档

0
0
约6.56千字
约 33页
2024-11-02 发布于湖北
举报
版权申诉
保障服务

两步聚类BIRCH算法.pptx

1、本文档共33页，可阅读全部内容。
2、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

BIRCH:

BalancedIterativeReducingandClusteringusingHierarchiesTianZhang,RaghuRamakrishnan,MironLivnyPresentedbyZhaoLi2023,Spring

12December20232OutlineIntroductiontoClusteringMainTechniquesinClusteringHybridAlgorithm:BIRCHExampleoftheBIRCHAlgorithmExperimentalresultsConclusions

Clustering

IntroductionDataclusteringconcernshowtogroupasetofobjectsbasedontheirsimilarityofattributesand/ortheirproximityinthevectorspace.MainmethodsPartitioning:K-Means…Hierarchical:BIRCH,ROCK,…Density-based:DBSCAN,…Agoodclusteringmethodwillproducehighqualityclusterswithhighintra-classsimilaritylowinter-classsimilarity12December20233

MainTechniques(1)

PartitioningClustering(K-Means)

step.112December20234initialcenterinitialcenterinitialcenter

K-MeansExample

Step.2

12December20235xxxnewcenterafter1stiterationnewcenterafter1stiterationnewcenterafter1stiteration

K-MeansExample

Step.3

12December20236newcenterafter2nditerationnewcenterafter2nditerationnewcenterafter2nditeration

MainTechniques(2)

HierarchicalClustering12December20237Multilevelclustering:level1hasnclusters?levelnhasonecluster,orupsidedown.AgglomerativeHC:startswithsingletonandmergeclusters(bottom-up).DivisiveHC:startswithonesampleandsplitclusters(top-down).Dendrogram

AgglomerativeHCExample12December20238NearestNeighborLevel2,k=7clusters.

12December20239NearestNeighbor,Level3,k=6clusters.

12December202310NearestNeighbor,Level4,k=5clusters.

12December202311NearestNeighbor,Level5,k=4clusters.

12December202312NearestNeighbor,Level6,k=3clusters.

12December202313NearestNeighbor,Level7,k=2clusters.

12December202314NearestNeighbor,Level8,k=1cluster.

Remarks12December202315PartitioningClusteringHierarchicalClusteringTimeComplexityO(n)O(n2logn)ProsEasytouseandRelativelyefficientOutputsadendrogramthatisdesi