Aiming at the shortcomings of DBSCAN ( Density-Based Spatial Clustering of Applications with Noise) ,this paper presents the concept of the attribute dimension partition by integrating the domain knowledge with thepartition idea.Then,the principles of the cluster merging and the pruning computation are demonstrated.Finally,an optimization method of DBSCAN is put forward based on the cloud computing programming model MapReduce,and the optimization method is verified through the data clustering of a real road transport information system.It isfound that the dataset partition helps to perform the concurrent computation,and the proposed optimization methodis superior to common statistical methods.