Estudio de estrategias para la aceleración de la convergencia del clustering mediante fuzzy c-means

En el presente trabajo de grado se lleva a cabo el estudio de estrategias para acelerar la convergencia del clustering empleando Fuzzy C-means (FCM), con la finalidad de encontrar la mejor variante del algoritmo de clustering FCM que permita reducir el tiempo de cómputo, coste computacional y calida...

Full description

Autores:
Trujillo Ortiz, Diana Carolina
Sacristan Hernandez, Carlos Serjeif
Tipo de recurso:
http://purl.org/coar/version/c_b1a7d7d4d402bcce
Fecha de publicación:
2016
Institución:
Universidad Industrial de Santander
Repositorio:
Repositorio UIS
Idioma:
spa
OAI Identifier:
oai:noesis.uis.edu.co:20.500.14071/35068
Acceso en línea:
https://noesis.uis.edu.co/handle/20.500.14071/35068
https://noesis.uis.edu.co
Palabra clave:
Cluster
Clustering
Fuzzy C-Means (Fcm)
Partición Difusa
Centroides
Convergencia.
In the present work
we perform a comparative study of strategies to accelerate the convergence of clustering using Fuzzy C-means (FCM)
in order to find the best implementation of the FCM clustering algorithm in terms of computational cost. The studied FCM variants related to the acceleration of clustering were Online Fuzzy C-means (OFCM)
and Single Pass Fuzzy C-means (SPFCM)
and Ramdom Sampling plus Extension Fuzzy C-means (RSEFCM) which are compared against the original FCM algorithm using three different databases in order to collect the necessary information about each clustering algorithm. I our experiments for assessing the speed up with respect to the classic FCM
we found that RSEFCM had the highest performance. With respect to the quality of the obtained partitions
RSEFCM yielded the more accurate and robust results among the studied versions of FCM. As for the complexity of the algorithms
experimental tests showed a linear behavior in the computational cost.
Rights
License
Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)