Improvements of DBSCAN Algorithm Based on Grid

In view of the low accuracy of boundary point clustered and the excessively high time complexity of DBSCAN algorithm, a DBSCAN algorithm is proposed based on GO-DBSCAN algorithm and OPTICS algorithm. The algorithm introduced the minimum reachable distance of OPTICS algorithm into the DBSCAN algorithm. Generally, the minimum distance between object and the near core object is a mainly considered factor for the clustering based on DBSCAN algorithm. Therefore, the introduce improves the clustering accuracy. The idea of grid query is proposed to reduce the size of traversal data of objects and this results in the decrease of DBSCAN algorithm'stime complexity. Before clustering, data set is divided into different grid based on grid query rules to reduce clustering time complexity. When project neighborhood is to be queried, it is necessary to traverse only the project near grid data other than the entire data set. The theoretical analysis and simulation results show that GO-DBSCAN can effectively improve the accuracy of boundary point clustered and reduce the time complexity.