Registro Completo |
Biblioteca(s): |
Embrapa Agricultura Digital. |
Data corrente: |
16/02/1998 |
Data da última atualização: |
21/01/2013 |
Autoria: |
CAMPAGNOLO, M. L.; COELHO, H.; CAPELO, J. H. |
Título: |
Knowledge based clustering of partially characterized objects. |
Ano de publicação: |
1995 |
Fonte/Imprenta: |
In: BRAZILIAN SYMPOSIUM ON ARTIFICIAL INTELLIGENCE, 12., 1995, Campinas. Advances in artificial intelligence: proceedings. Berlin: Springer, 1995. |
Páginas: |
p. 161-170. |
Série: |
(Lecture Notes in Artificial Intelligence, 991; Lecture Notes in Computer Science). |
ISBN: |
3-540-60436-7 |
Idioma: |
Inglês |
Notas: |
SBIA'95. Ed. by Jacques Wainer and Ariadne Carvalho. |
Conteúdo: |
Cluster analysis can be generally described as a way to discover and identify concepts from a set of multicharacterized objects. In cluster analysis, each concept is described extensionally by a subset of objects called a class. Taking those concepts as reference, it is known, when the characteristics are not all available, that the cluster analysis procedure can lead to spurious class definitions, caused by the loss of information on the object descriptions. So, one of the possible techniques to handle this problem consists of incorporating more knowledge about the structure of classes of objects in this classification process. In this paper we propose and defend a new method for clustering partially characterized objects, by combining three approaches: 1) the use of probability theory to represent uncertainty on objects' description, 2) the adoption of a Bayesian network model to support and explore knowledge on dependence relations among the objects' descriptive variables, and 3) the introduction of a new similarity measure consistent with the used objects' representation. We also apply this method to a vegetation classification problem and we show that the achieved classification is less sensitive to data erosion than classifications obtained just from the known characteristics of the objects. |
Palavras-Chave: |
Análise de cluster; Inteligência artificial. |
Thesaurus Nal: |
artificial intelligence; Cluster analysis. |
Categoria do assunto: |
X Pesquisa, Tecnologia e Engenharia |
Marc: |
LEADER 02188naa a2200241 a 4500 001 1005962 005 2013-01-21 008 1995 bl uuuu u00u1 u #d 020 $a3-540-60436-7 100 1 $aCAMPAGNOLO, M. L. 245 $aKnowledge based clustering of partially characterized objects. 260 $c1995 300 $ap. 161-170. 490 $a(Lecture Notes in Artificial Intelligence, 991; Lecture Notes in Computer Science). 500 $aSBIA'95. Ed. by Jacques Wainer and Ariadne Carvalho. 520 $aCluster analysis can be generally described as a way to discover and identify concepts from a set of multicharacterized objects. In cluster analysis, each concept is described extensionally by a subset of objects called a class. Taking those concepts as reference, it is known, when the characteristics are not all available, that the cluster analysis procedure can lead to spurious class definitions, caused by the loss of information on the object descriptions. So, one of the possible techniques to handle this problem consists of incorporating more knowledge about the structure of classes of objects in this classification process. In this paper we propose and defend a new method for clustering partially characterized objects, by combining three approaches: 1) the use of probability theory to represent uncertainty on objects' description, 2) the adoption of a Bayesian network model to support and explore knowledge on dependence relations among the objects' descriptive variables, and 3) the introduction of a new similarity measure consistent with the used objects' representation. We also apply this method to a vegetation classification problem and we show that the achieved classification is less sensitive to data erosion than classifications obtained just from the known characteristics of the objects. 650 $aartificial intelligence 650 $aCluster analysis 653 $aAnálise de cluster 653 $aInteligência artificial 700 1 $aCOELHO, H. 700 1 $aCAPELO, J. H. 773 $tIn: BRAZILIAN SYMPOSIUM ON ARTIFICIAL INTELLIGENCE, 12., 1995, Campinas. Advances in artificial intelligence: proceedings. Berlin: Springer, 1995.
Download
Esconder MarcMostrar Marc Completo |
Registro original: |
Embrapa Agricultura Digital (CNPTIA) |
|