云环境下并行DBSCAN聚类算法研究

Research on Parallel DBSCAN Clustering Algorithm in Cloud Environment

DBSCAN算法是一种基于密度的快速聚类算法,虽然在处理大规模数据时可以发现其中的噪声数据,但聚类效率不高,输入/输出消耗大,聚类结果准确率低。本文在云计算平台Hadoop环境下,将MapReduce编程模型的高并行性引入该算法,设计出一种并行DBSCAN算法,提高传统DBSCAN算法的执行效率,通过对比实验结果证明了该算法聚类的准确性和时效性。

DBSCAN algorithm is a density-based fast clustering algorithm. Although the noise data can be found when dealing with large-scale data,the clustering efficiency is not high,the input/output consumption is large and the accuracy of clustering results is low. In this paper,the parallelism of the MapReduce programming model is introduced into the Hadoop environment,and a parallel DBSCAN algorithm is designed to improve the efficiency of the traditional DBSCAN algorithm. The accuracy of the algorithm is proved by comparing the experimental results and timeliness.