Design and optimization of DBSCAN Algorithm based on CUDA

DBSCAN is a very classic algorithm for data clustering, which is widely used in many fields. However, with the data scale growing much more bigger than before, the traditional serial algorithm can not meet the performance requirement. Recently, parallel computing based on CUDA has developed very fast and has great advantage on big data. This paper summarizes the algorithms proposed before and improves the performance of the old DBSCAN algorithm by CUDA. The algorithm uses shared memory as much as possible compared with other algorithms and it has very good scalability. A data set is tested on the algorithm of new version. Finally, we analyze the results and give a conclusion that our algorithm is approximately 97 times faster than the serial version.