Estudio de pertinencia de algoritmos en procesos de descubrimiento de reglas de pertenencia a grupos
Files
Date
2019-02-20
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Escuela de Posgrado - Facultad Regional Buenos Aires
Abstract
En el campo de la Ingeniería de explotación de información, el proceso de descubrimiento de
reglas de pertenencia a grupos se caracteriza por la utilización combinada de un proceso de
descubrimiento de grupos (clustering) y uno de inducción de reglas. Dada la variedad de
algoritmos de clustering e inducción de reglas disponibles en la actualidad, es de interés poder
conocer a priori que pareja de algoritmos es más conveniente para un set de datos, en a base sus
características. En esta tesis, se propone un proceso que permite validar el rendimiento de los
algoritmos, en base a métricas internas, para distintos tipos de sets de datos, con características
específicas, de forma tal que permita comprender bajo que características cada pareja de
algoritmos ofrece mejor rendimiento.
In the data mining field, the group membership rules discovery process consists of utilizing a group discovery (clustering) and rules induction process combined. Due to the broad variety of clustering and rules induction algorithms currently available, it is considered of interest to know beforehand which pair of algorithms is more convenient for a given dataset, based just on its properties. In this thesis, a process that allows to validate algorithms performance, based on internal metrics, for different datasets, with specific characteristics, is proposed, so that it allows to understand under which characteristics, each algorithm pair offers better performance.
In the data mining field, the group membership rules discovery process consists of utilizing a group discovery (clustering) and rules induction process combined. Due to the broad variety of clustering and rules induction algorithms currently available, it is considered of interest to know beforehand which pair of algorithms is more convenient for a given dataset, based just on its properties. In this thesis, a process that allows to validate algorithms performance, based on internal metrics, for different datasets, with specific characteristics, is proposed, so that it allows to understand under which characteristics, each algorithm pair offers better performance.
Description
Keywords
Sistemas de Información, Minería de datos, Clustering, Inducción de reglas, Descubrimiento de reglas de pertenencia a grupos, Estudio de algoritmos, Data mining, Rules induction, Membership rules discovery, Algorithms study
Citation
Endorsement
Review
Supplemented By
Referenced By
Creative Commons license
Except where otherwised noted, this item's license is described as info:eu-repo/semantics/openAccess