Multivariate Clustering Procedures with Variable Metrics

Maronna, Ricardo A.; Jacovkis, Pablo M.

doi:10.2307/2529203

Cited by 51 publications

(30 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The untrimmed case alpha = 0, restr = "deter" and restr.fact = 1 was already outlined in Maronna and Jacovkis (1974), as the only sensible way to avoid (Mahalanobis distance modified) k-means type algorithms to return clusters of a few almost collinear observations. The possibility of trimming data was also considered in Gallegos (2002) who implicitly assumed |Σ 1 | = · · · = |Σ k | (and so restr.fact = 1).…”

Section: Constraints On the Determinantsmentioning

confidence: 99%

tclust: AnRPackage for a Trimming Approach to Cluster Analysis

Fritz¹,

García-Escudero²,

Mayo-Íscar³

2012

J. Stat. Soft.

View full text Add to dashboard Cite

Outlying data can heavily influence standard clustering methods. At the same time, clustering principles can be useful when robustifying statistical procedures. These two reasons motivate the development of feasible robust model-based clustering approaches. With this in mind, an R package for performing non-hierarchical robust clustering, called tclust, is presented here. Instead of trying to "fit" noisy data, a proportion α of the most outlying observations is trimmed. The tclust package efficiently handles different cluster scatter constraints. Graphical exploratory tools are also provided to help the user make sensible choices for the trimming proportion as well as the number of clusters to search for.

show abstract

Section: Constraints On the Determinantsmentioning

confidence: 99%

tclust: AnRPackage for a Trimming Approach to Cluster Analysis

Fritz¹,

García-Escudero²,

Mayo-Íscar³

2012

J. Stat. Soft.

View full text Add to dashboard Cite

show abstract

“…O primeiro conjunto (DS1) contempla nove instâncias conhecidas da literatura com a quantidade de objetos entre 75 e 1484 e dimensões (quantidade de atributos) entre 2 e 13 [Fisher 1936] [Ruspini 1970] [Maronna and Jacovkis 1974] [Hastie et al 2001] [Naldi 2011] [Wang 2007].…”

Section: -Instâncias Utilizadasunclassified

Uma Heurística Baseada Em Densidade Para O Problema De Agrupamento Automático

Semaan¹,

Fadel²,

Brito³

et al. 2016

Anais Do XVIII Simpósio De Pesquisa Operacional &Amp; Logística Da Marinha

View full text Add to dashboard Cite

ResumoA análise de agrupamentos agrega vários métodos que visam identificar grupos dentro de um conjunto de dados. Este artigo apresenta novas heurísticas baseadas na metaheurística Busca Local Iterada para resolver o Problema de Agrupamento Automático, qual seja o problema de determinar o número ideal de grupos para uma base dados. Para tal, em uma das fases da aplicação desta heurística, foi utilizado o índice silhueta, que combina conceitos de coesão e separação e é considerado pelas heurísticas propostas para avaliar a qualidade das soluções. De acordo com os experimentos computacionais reportados neste trabalho, verifica-se que a nova heurística ILS-DBSCAN é muito eficiente no que concerne ao tempo de processamento e muito eficaz quanto à qualidade das soluções obtidas, quando comparado com outros métodos da literatura. Em geral, os resultados desta nova heurística foram superiores aos resultados relatados na literatura. Dessa maneira o ILS-DBSCAN apresenta-se como um algoritmo promissor para a resolução do problema abordado. Palavras-chave:

show abstract

“…In some clustering applications, clusters might assume more of an elliptical shape, perhaps with different spatial orientations and sizes. Alternative criteria for various shapes and orientations, which are principally based on determinants of within-cluster sums of squares matrices and/or between-cluster sums of squares matrices have been proposed by a number of authors (Banfield & Raftery, 1993;Friedman & Rubin, 1967;Maronna & Jacovkis, 1974;Marriott, 1982;Scott & Symons, 1971;Symons, 1981;Windham, 1987).…”

Section: Critiquing Alternative Clustering Criteriamentioning

confidence: 99%

Measuring Human Performance on Clustering Problems: Some Potential Objective Criteria and Experimental Research Opportunities

Brusco¹

2007

The Journal of Problem Solving

View full text Add to dashboard Cite

The study of human performance on discrete optimization problems has a considerable history that spans various disciplines. The purpose of this paper is to outline a program of study for the measurement of human performance on discrete optimization problems related to clustering of points in the two-dimensional plane. I describe possible objective criteria for clustering problems, the measurement of agreement of solutions produced by subjects, and categories of experiments for investigating human performance on clustering problems. To facilitate future experimental testing of human subjects on clustering problems, optimal partitions were obtained for 233 two-dimensional clustering problems ranging in size from 10 to 70 points. For each test problem, an optimal solution was obtained for each of three objective criteria: (a) maximizing partition split, (b) minimizing partition diameter, and (c) minimizing within-cluster sums of squares, and similarity of the solutions among these criteria has been computed.

show abstract

Multivariate Clustering Procedures with Variable Metrics

Cited by 51 publications

References 8 publications

tclust: AnRPackage for a Trimming Approach to Cluster Analysis

tclust: AnRPackage for a Trimming Approach to Cluster Analysis

Uma Heurística Baseada Em Densidade Para O Problema De Agrupamento Automático

Measuring Human Performance on Clustering Problems: Some Potential Objective Criteria and Experimental Research Opportunities

Contact Info

Product

Resources

About