Scalable model-based cluster analysis using clustering features

Huidong JIN; Kwong Sak LEUNG; Man Leung WONG; Zong Ben XU

doi:10.1016/j.patcog.2004.07.012

Scalable model-based cluster analysis using clustering features

Huidong JIN, Kwong Sak LEUNG, Man Leung WONG, Zong Ben XU

Department of Computing and Decision Sciences

Research output: Journal Publications › Journal Article (refereed) › peer-review

23 Citations (Scopus)

Abstract

We present two scalable model-based clustering systems based on a Gaussian mixture model with independent attributes within clusters. They first summarize data into sub-clusters, and then generate Gaussian mixtures from their clustering features using a new algorithm - EMACF. EMACF approximates the aggregate behavior of each sub-cluster of data items in the Gaussian mixture model. It provably converges. The experiments show that our clustering systems run one or two orders of magnitude faster than the traditional EM algorithm with few losses of accuracy.

Original language	English
Pages (from-to)	637-649
Number of pages	13
Journal	Pattern Recognition
Volume	38
Issue number	5
DOIs	https://doi.org/10.1016/j.patcog.2004.07.012
Publication status	Published - 1 May 2005

Bibliographical note

The authors appreciate the anonymous referees for their valuable comments to strengthen the paper, and T. Zhang, R. Ramakrishnan, M. Livny, and V. Ganti for the BIRCH source code.

Funding

The work was partially supported by RGC Grant CUHK 4212/01E of Hong Kong, Lingnan University direct grant (RES-021/200), RGC Grant LU 3009/02E of Hong Kong, and the Nature Science Foundation Project (No. 10371097) of China.

Keywords

Cluster analysis
Clustering feature
Convergence
Data mining
Expectation maximization
Gaussian mixture model
Scalable

Access to Document

10.1016/j.patcog.2004.07.012

Cite this

@article{01ec27d183e5411b91423784eec5a7f8,

title = "Scalable model-based cluster analysis using clustering features",

abstract = "We present two scalable model-based clustering systems based on a Gaussian mixture model with independent attributes within clusters. They first summarize data into sub-clusters, and then generate Gaussian mixtures from their clustering features using a new algorithm - EMACF. EMACF approximates the aggregate behavior of each sub-cluster of data items in the Gaussian mixture model. It provably converges. The experiments show that our clustering systems run one or two orders of magnitude faster than the traditional EM algorithm with few losses of accuracy.",

keywords = "Cluster analysis, Clustering feature, Convergence, Data mining, Expectation maximization, Gaussian mixture model, Scalable",

author = "Huidong JIN and LEUNG, {Kwong Sak} and WONG, {Man Leung} and XU, {Zong Ben}",

note = "The authors appreciate the anonymous referees for their valuable comments to strengthen the paper, and T. Zhang, R. Ramakrishnan, M. Livny, and V. Ganti for the BIRCH source code.",

year = "2005",

month = may,

day = "1",

doi = "10.1016/j.patcog.2004.07.012",

language = "English",

volume = "38",

pages = "637--649",

journal = "Pattern Recognition",

issn = "0031-3203",

publisher = "Elsevier Ltd",

number = "5",

}

TY - JOUR

T1 - Scalable model-based cluster analysis using clustering features

AU - JIN, Huidong

AU - LEUNG, Kwong Sak

AU - WONG, Man Leung

AU - XU, Zong Ben

N1 - The authors appreciate the anonymous referees for their valuable comments to strengthen the paper, and T. Zhang, R. Ramakrishnan, M. Livny, and V. Ganti for the BIRCH source code.

PY - 2005/5/1

Y1 - 2005/5/1

N2 - We present two scalable model-based clustering systems based on a Gaussian mixture model with independent attributes within clusters. They first summarize data into sub-clusters, and then generate Gaussian mixtures from their clustering features using a new algorithm - EMACF. EMACF approximates the aggregate behavior of each sub-cluster of data items in the Gaussian mixture model. It provably converges. The experiments show that our clustering systems run one or two orders of magnitude faster than the traditional EM algorithm with few losses of accuracy.

AB - We present two scalable model-based clustering systems based on a Gaussian mixture model with independent attributes within clusters. They first summarize data into sub-clusters, and then generate Gaussian mixtures from their clustering features using a new algorithm - EMACF. EMACF approximates the aggregate behavior of each sub-cluster of data items in the Gaussian mixture model. It provably converges. The experiments show that our clustering systems run one or two orders of magnitude faster than the traditional EM algorithm with few losses of accuracy.

KW - Cluster analysis

KW - Clustering feature

KW - Convergence

KW - Data mining

KW - Expectation maximization

KW - Gaussian mixture model

KW - Scalable

UR - http://commons.ln.edu.hk/sw_master/2385

UR - http://www.scopus.com/inward/record.url?scp=13644262819&partnerID=8YFLogxK

U2 - 10.1016/j.patcog.2004.07.012

DO - 10.1016/j.patcog.2004.07.012

M3 - Journal Article (refereed)

SN - 0031-3203

VL - 38

SP - 637

EP - 649

JO - Pattern Recognition

JF - Pattern Recognition

IS - 5

ER -

Scalable model-based cluster analysis using clustering features

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this