Path Clustering: Grouping in an Efficient Way Complex Data Distributions
R. Q. A. Fernandes1, W. A. Pinheiro1,2,3, G. B. Xexéo3, J. M. de Souza3
1 Centro de Desenvolvimento de Sistemas, SMU, Brasília, DF, CEP, Brazil
2 Instituto Militar de Engenharia, Praia Vermelha, Urca, Rio de Janeiro, RJ, CEP, Brazil
3 COPPE/UFRJ, Universidade Federal do Rio de Janeiro, RJ, PO Box 68.501, Brazil
E-mail: ricardo@cds.eb.mil.br, awallace@cos.ufrj.br, xexeo@cos.ufrj.br, jano@cos.ufrj.br
Abstract: This work proposes an algorithm that uses paths based on tile segmentation to build complex clusters. After allocating data items (points) to geometric shapes in tile format, the complexity of our algorithm is related to the number of tiles instead of the number of points. The main novelty is the way our algorithm goes through the grids, saving time and providing good results. It does not demand any configuration parameters from users, making easier to use than other strategies. Besides, the algorithm does not create overlapping clusters, which simplifies the interpretation of results.
Research Paper Link: http://dspace.chitkara.edu.in/jspui/bitstream/123456789/703/3/4-%20Path%20Clustering%20in%20a%20efficient%20way%20complex%20data%20-%20Fernandes.pdf
DOI: https://doi.org/10.15415/jotitt.2017.52004
1 Centro de Desenvolvimento de Sistemas, SMU, Brasília, DF, CEP, Brazil
2 Instituto Militar de Engenharia, Praia Vermelha, Urca, Rio de Janeiro, RJ, CEP, Brazil
3 COPPE/UFRJ, Universidade Federal do Rio de Janeiro, RJ, PO Box 68.501, Brazil
E-mail: ricardo@cds.eb.mil.br, awallace@cos.ufrj.br, xexeo@cos.ufrj.br, jano@cos.ufrj.br
Abstract: This work proposes an algorithm that uses paths based on tile segmentation to build complex clusters. After allocating data items (points) to geometric shapes in tile format, the complexity of our algorithm is related to the number of tiles instead of the number of points. The main novelty is the way our algorithm goes through the grids, saving time and providing good results. It does not demand any configuration parameters from users, making easier to use than other strategies. Besides, the algorithm does not create overlapping clusters, which simplifies the interpretation of results.
Research Paper Link: http://dspace.chitkara.edu.in/jspui/bitstream/123456789/703/3/4-%20Path%20Clustering%20in%20a%20efficient%20way%20complex%20data%20-%20Fernandes.pdf
DOI: https://doi.org/10.15415/jotitt.2017.52004
Comments
Post a Comment