Abstract—Abstract—For it can identify the clusters with any shape and tackle the boundary points effectively, the typical density-based method of DBSCAN was widely applied to the clustering analysis. But the algorithm still has some shortcomings, such as the high time complexity, clustering effect is very dependent on the initial value of the parameter, and the low accuracy of the boundary points tackling processes. This paper put forward GO-DBSCAN, which based on the DBSCAN and OPTICS algorithms. GO-DBSCAN improved the accuracy while processing the boundary points, that cause of it introduced the minimum acceptable distance of OPTICS. In order to reduce the time complexity of clustering processes, it also proposed the method of grid-based query while it retraverse the neighborhood. At the end of this paper, we proved that GO-DBSCAN would perform better both on the accuracy of boundary points processing and time complexity.
Index Terms—Index Terms—Clustering, DBSCAN, OPTICS, grid, boundary point.
Ling Feng, Kejian Liu, and Fuxi Tang are with the Xihua University, Sichuan, China (e-mail: lyn_ling_feng@163.com, liukejian@gmail.com, Fuxi_tang@163.com).
Qingrui Meng is with Tibet Feiyue Intelligence Technology CO., Ltd, China (e-mail: 414893358@qq.com).
[PDF]
Cite:Ling Feng, Kejian Liu, Fuxi Tang, and Qingrui Meng, "GO-DBSCAN: Improvements of DBSCAN Algorithm Based on Grid," International Journal of Computer Theory and Engineering vol. 9, no. 3, pp. 151-155, 2017.
