Publicação
A clustering algorithm based on fitness probability scores for cluster centers optimization
| Resumo: | In the present paper, we propose an iterative clustering approach that sequentially applies five processes, namely: the assign, delete, split, delete and optimization. It is based on the fitness probability scores of the cluster centers to identify the least fitted centers to undergo an optimization process, aiming to improve the centers from one iteration to another. Moreover, the parameters of the algorithm for the delete, split and optimization processes are dynamically tuned as problem dependent functions. The presented clustering algorithm is evaluated using four data sets, two randomly generated and two well-known sets. The obtained clustering algorithm is compared with other clustering algorithms through the visualization of the clustering, the value of a validity measure and the value of the objective function of the optimization process. The comparison of results shows that the proposed clustering algorithm is effective and robust. |
|---|---|
| Autores principais: | Costa, M. Fernanda P. |
| Outros Autores: | Rocha, Ana Maria A. C.; Fernandes, Edite M. G. P. |
| Assunto: | Clustering analysis Fitness probability score Differential evolution |
| Ano: | 2021 |
| País: | Portugal |
| Tipo de documento: | comunicação em conferência |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | In the present paper, we propose an iterative clustering approach that sequentially applies five processes, namely: the assign, delete, split, delete and optimization. It is based on the fitness probability scores of the cluster centers to identify the least fitted centers to undergo an optimization process, aiming to improve the centers from one iteration to another. Moreover, the parameters of the algorithm for the delete, split and optimization processes are dynamically tuned as problem dependent functions. The presented clustering algorithm is evaluated using four data sets, two randomly generated and two well-known sets. The obtained clustering algorithm is compared with other clustering algorithms through the visualization of the clustering, the value of a validity measure and the value of the objective function of the optimization process. The comparison of results shows that the proposed clustering algorithm is effective and robust. |
|---|