Study Of Spatial Data Mining Algorithm Based On Density Clustering

As the rapid development of the spatial data obtaining, the number and complexity of spatial data increase rapidly, which has far exceeded the capacity of people's interpretation result to"spatial data explosion but knowledge poor". So it is very necessary for making use of spatial data mining and knowledge discovery to mine aforehand unknown and potentially useful spatial patterns from spatial database. For spatial data mining, the complexity and large number of spatial data make the spatial data mining difficult. How to mine useful information from spatial database with the data mining method is a major problem that needs to be solved currently.According to the intensively researching on the status of spatial data mining algorithm, comparing and analyzing the existing spatial data-mining algorithm. This paper discusses the spatial index technology that is the influencing factor of spatial data mining. Aiming at time complexity and parameter sensitivity bases on the density clustering algorithm for the space density of data mining in dealing with large number of data points, the RCDBSCAN (Restricted Candidate-points with the vector based on DBSCAN) algorithm has been put forward in this paper. This algorithm bases on traditional DBSCAN algorithm, and regards finding linear cluster structure as the goal. The RCDBSCAN algorithm adopts restricted candidates-point with vector technology to select candidate-points, and reduces the data points in processing. At the same time, data-partitioning method has been used to divide data sets and make density-based clustering in each subset in RCDBSCAN algorithm, which reduces the sensitivity of the initialized parameter. Furthermore, in order to accelerate the data access, the SS index technology has been applied to the algorithm.Finally,the validity of RCDBSCAN algorithm has been confirmed by the experiments on certain area terrain data of TaiHang mountain, which is provided by states geography institute. The results show that the RCDBSCAN algorithm reduces the time-complexity greatly than traditional DBSCAN algorithm, and receives good effect in processing of the boundary points. Therefore, the RCDBSCAN algorithm can provides good guarantee for 3D terrain analysis.