CGDBSCAN: DBSCAN Algorithm Based on Contribution and Grid

GbDBSCAN (an efficient grid-based DBSCAN algorithm) is an excellent improved DBSCAN algorithm, which makes up the defects that DBSCAN algorithm is sensitive to clustering parameters and unable to deal with large database, and retains the advantage of separating noise and finding arbitrary shape clusters. However, in GbDBSCAN, the grid technique treats the total number of points in one grid as the grid dense, and this simple treatment will depress the clustering accuracy. Therefore, CGDBSCAN is proposed in this paper, and within it ´migration-coefficient´ conception is presented firstly. With the optimization effect of contribution and migration-coefficient, the optimal selection of parameter Eps and the efficient SP-tree query index, the accuracy of clustering result is effectively improved while ensuring the operational efficiency of this algorithm.