Algoritmos conceptuales restringidos basados en semillas
The non-supervised classification algorithms determine clusters such that objects in the same cluster are very similar among them, while objects in different clusters are not similar. However, there are some problems where it is required, besides determining the clusters, to know the properties that...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2007 |
| País: | México |
| Institución: | Instituto Nacional de Astrofísica, Óptica y Electrónica |
| Repositorio: | Repositorio Institucional del INAOE |
| Idioma: | español |
| OAI Identifier: | oai:inaoe.repositorioinstitucional.mx:1009/581 |
| Acceso en línea: | http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/581 |
| Access Level: | acceso abierto |
| Palabra clave: | info:eu-repo/classification/Teoría del algoritmo/Algorithm theory info:eu-repo/classification/Herramientas de clúster/Cluster tools info:eu-repo/classification/Clasificación/Classification info:eu-repo/classification/cti/7 info:eu-repo/classification/cti/33 info:eu-repo/classification/cti/3304 info:eu-repo/classification/cti/120302 |
| Sumario: | The non-supervised classification algorithms determine clusters such that objects in the same cluster are very similar among them, while objects in different clusters are not similar. However, there are some problems where it is required, besides determining the clusters, to know the properties that characterize them. This problem is known as conceptual clustering. There are different methods that allow to solve the conceptual clustering problem, one of them is the conceptual k-means algorithm, which is a conceptual version of the k-means algorithm; one of the most studied and used algorithms for solving the restricted non-supervised classification problem (when the number of clusters is specified a priori). The main characteristic of the conceptual k-means algorithm is that it requires generalization lattices for the construction of the concepts. The generalization lattices for the qualitative features must be given and the generalization lattices for the quantitative features are built starting from a codification of their values. In this thesis, an improvement of the conceptual k-means algorithm, which uses a different strategy for building the clusters and in the characterization phase, a different generalization lattice for the quantitative features, is proposed. The inconvenience of using generalization lattices is that, in general, it is difficult to determine the generalization lattices. Also, there are not automatic methods to build the generalization lattices; therefore, this task must be done by the user. For this reason, in this thesis, a conceptual k-means algorithm that does not depend on generalization lattices for building the concepts is proposed. Finally, in this thesis, two fuzzy conceptual clustering algorithms, which are fuzzy versions of the proposed hard conceptual clustering algorithms, are proposed. |
|---|