VDBSCAN plus : Performance Optimization Based on GPU Parallelism

• Autor(a) principal: Valencio, Carlos Roberto [UNESP]
• Data de Publicação: 2013
• Outros Autores: Daniel, Guilherme Priolli [UNESP], Medeiros, Camila Alves de [UNESP], Cansian, Adriano Mauro [UNESP], Baida, Luiz Carlos [UNESP], Ferrari, Fernando [UNESP], Horng, S. J.
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1109/PDCAT.2013.11; http://hdl.handle.net/11449/197449
• topic: spatial data mining; spatial clustering; GPU (Graphics Processing Unit); VDBSCAN (Varied Density Based Spatial Clustering of Applications with Noise)

Spatial data mining techniques enable the knowledge extraction from spatial databases. However, the high computational cost and the complexity of algorithms are some of the main problems in this area. This work proposes a new algorithm referred to as VDBSCAN+, which derived from the algorithm VDBSCAN (Varied Density Based Spatial Clustering of Applications with Noise) and focuses on the use of parallelism techniques in GPU (Graphics Processing Unit), obtaining a significant performance improvement, by increasing the runtime by 95% in comparison with VDBSCAN.