A density-based cluster validity approach using multi-representatives

Halkidi, Maria; Vazirgiannis, Michalis

doi:10.1016/j.patrec.2007.12.011

Cited by 113 publications

(78 citation statements)

References 21 publications

(21 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…As mentioned earlier, the CDbw clustering validity index which allows to incorporate inter-cluster and intra-cluster density information is used in conjunction with two spatial information measures to determine which cluster pair to be merged. Although the CDbw index has a reasonable computational complexity in comparison with others clustering validity indexes (Halkidi & Vazirgiannis, 2002), the local use of it to decide about the cluster fusion added to the large data volume that normally arises from remote sensing images may require a considerable processing time for clustering the SOM. For this reason, seeking to reduce the processing volume, instead of calculating the CDbw index directly on image original data the method proposed here performs the calculation of it using the own SOM prototype vectors (which represent the image original data).…”

Section: Hierarchical Clustering Methodsmentioning

confidence: 99%

“…Seeking to overcome the limitations of these approaches, Wu & Chow (2004) proposed then a hierarchical algorithm for clustering the SOM that uses more information about the data in each cluster in addition to inter-cluster distances. To determine which pair of clusters to be merged the proposed algorithm locally uses the CDbw clustering validity index (Composing Density Between and Within Clusters), presented in Halkidi & Vazirgiannis (2002), which allows to incorporate the inter-cluster and intra-cluster density into merging criteria in addition to distance information. Wu and Chow showed that the algorithm proposed by them clusters data better than the classical clustering algorithms on the SOM.…”

Section: Second Level Clustering -Segmentation Of the Sommentioning

confidence: 99%

See 1 more Smart Citation

Land-Cover Classification Using Self-Organizing Maps Clustered with Spectral and Spatial Information

Gonçalves¹,

Costa²,

Netto³

2011

Self Organizing Maps - Applications and Novel Algorithm Design

View full text Add to dashboard Cite

Section: Hierarchical Clustering Methodsmentioning

confidence: 99%

Section: Second Level Clustering -Segmentation Of the Sommentioning

confidence: 99%

Land-Cover Classification Using Self-Organizing Maps Clustered with Spectral and Spatial Information

Gonçalves¹,

Costa²,

Netto³

2011

Self Organizing Maps - Applications and Novel Algorithm Design

View full text Add to dashboard Cite

“…An improved version of Clustering Using Representative is proposed in [17] to pick different numbers of representatives for a variety of clusters based on cluster density. A multi-representative approach based on density is planned in [18]. In the fuzzy clustering approach called fuzzy clustering with weighted prototype [19], every cluster is characterized by a variety of weighted medoids.…”

Section: Related Workmentioning

confidence: 99%

“…In the fuzzy clustering approach called fuzzy clustering with weighted prototype [19], every cluster is characterized by a variety of weighted medoids. Unlike from other multi representatives based methods [16] [18]. Where representatives of every cluster are choose to be a prespecified variety with equal weights, in Partitional Fuzzy Clustering, the weights as well as the variety of representative objects in every cluster are determined based on the nature of the dataset.…”

Section: Related Workmentioning

confidence: 99%

Evaluation of Clustering around Weighted Prototype and Genetic Algorithm for Document Categorization

Jain¹,

Shrivastava²

2015

IJCA

View full text Add to dashboard Cite

Document clustering is very important in the field of text categorization. Genetic algorithm, which is an optimization based technique which can be applied for finding out the best cluster centres easily by computing fitness values of data points. While clustering around weighted prototype technique is especially helpful when proper pairwise similarities are available. This technique does not find global solution of the objective function. Experimental result shows that F-measure and Normalized mutual information of genetic algorithm is better than clustering around weighted prototype for 20 Newsgroup dataset. F-measure and accuracy of genetic algorithm is better than clustering around weighted prototype for the Reuter-21578 dataset.

show abstract

“…Apesar do sucesso dos processos de agrupamento de dados, considera-se que resultados satisfatórios, a partir do ponto de vista de um determinado usuário, só serão atingidos se for possível avaliar o comportamento, importância ou influência dos dados e atributos nos resultados obtidos, bem como entender se os grupos são similares ou não similares (HALKIDI;VAZIRGIANNIS, 2002). Assim, desenvolvemos uma metodologia que possibilita a aplicação da metáfora compacta apresentada anteriormente na análise de agrupamentos.…”

Section: Analisando Agrupamentosunclassified

Uma nova metáfora visual escalável para dados tabulares e sua aplicação na análise de agrupamentos

Mosquera¹

View full text Add to dashboard Cite

Este trabalho é dedicado a minha mãe Marisol e meu Pai Jose, também a todas as pessoas que estão em procura da felicidade, todas aquelas que dia a dia superam as adversidade para conseguir o sucesso, Especialmente os pesquisadores, mergulhando-se no estudo e na busca da verdade. AGRADECIMENTOSAgradeço primeiramente a Deus e a todas as energias que convergiram para o sucesso. Ao meu orientador, Professor Dr. Fernando Viera Paulovich, a orientação, ajuda e pela paciência com minha desorganização, prometo melhorar mais do que já tenho melhorado. Nunca vou esquecer esse caminho no qual andei, Obrigado.Aos meus pais, Jose Cordoba e Marisol Mosquera, pelas palavras, pelo apoio que dão na minha vida e acreditar em mim. Obrigado Leidy Yohana, por me escutar e apoiar todos os dias durante o meu mestrado.Aos meus amigos que se preocuparam por mim, pelas alegrias que me proporcionaram quando eu voltava à casa.A todos os meus amigos que fiz aqui no Brasil. Especialmente no ICMC, Gladys, nunca vou esquecer de seu apoio, companheirismo quando precisei, muito obrigado. Filomen, colega e amigo de turma, foram muitas noites de estudo no laboratório, muito companheirismo e cooperação. Danilo e Tácito, obrigado pelas dicas e compartilhar palavras, tudo isso foi muito importante para eu poder me adaptar. Gabriel, muito obrigado pela ajuda na escrita, Paulo, Clausius, Fabio, Diego, Lucas, Nícolas, em fim a todos os colegas do laboratório de pesquisa em Visualização, Imagens e Computação. Muito Obrigado pela amizade, minhas palavras serão sempre de gratidão. Respeito muito as suas coragem na busca do sucesso. Ao instituto de Ciências Matemáticas e de Computação ICMC-USP, amigos de aulas, seus professore e funcionários.Ao governo do Chocó-Colômbia e ao Departamento Administrativo de Ciência, Tecnologia e Inovação -COLCIENCIAS, processo -acordo n ∘ 530/2014. A todos que de alguma forma contribuíram para a realização deste trabalho. Mi casa es su casa! "Os campeões não são feitos nas academias. Os campeões são feitos de algo que eles têm profundamente dentro de si -um desejo, um sonho, uma visão -" (Muhammad The rapid evolution of computing resources has enabled large datasets to be stored and retrieved. However, exploring, understanding and extracting useful information is still a challenge. Among the computational tools to address this problem, information visualization techniques enable the data analysis employing the human visual ability by making a graphic representation of the data set, and data mining provides automatic processes for the discovery and interpretation of patterns.Despite the recent popularity of information visualization methods, a recurring problem is the low visual scalability when analyzing large data sets resulting in context loss and visual disorder.To represent large datasets reducing the loss of relevant information, the process of aggregation is being used. Aggregation decreases the amount of data to be represented, preserving the distribution and trends of the original dataset. Regarding data mining, infor...

show abstract

A density-based cluster validity approach using multi-representatives

Cited by 113 publications

References 21 publications

Land-Cover Classification Using Self-Organizing Maps Clustered with Spectral and Spatial Information

Land-Cover Classification Using Self-Organizing Maps Clustered with Spectral and Spatial Information

Evaluation of Clustering around Weighted Prototype and Genetic Algorithm for Document Categorization

Uma nova metáfora visual escalável para dados tabulares e sua aplicação na análise de agrupamentos

Contact Info

Product

Resources

About