Research-based Obstacles And Direction Of The Density Of Bound Clustering Algorithm

With the development of database technology and the diversification of means for people accessing to data, the number, size and complexity of the spatial data are increasing dramatically, and these data are increasing geometrically or exponentially, which's far beyond the explanatory ability of people, Therefore, how to extract useful information efficiently from huge spatial data has become a problem studied widely in spatial data mining. Spatial clustering that is an important means of data mining has been widely used; however, traditional spatial clustering algorithms whose aim was to improve the algorithm performance are applied to solve large-scale data, and yet ignore obstacle constraints of data space. Only a few algorithms consider the effect of obstacles to clustering in the literature, however there are shortcomings in the algorithm efficiency, clustering quality and other aspects. Complicated terrain, different natural conditions and other issues in the highway routing bring a certain degree of difficulty to spatial clustering. The research on density clustering algorithm in obstacle space has become a hotspot of spatial data mining.Currently, most of the obstacle constraints-based spatial clustering algorithms are improved on the basis of the traditional clustering algorithms. As the added obstacle constraints, it makes the computation of the original algorithm increase dramatically, or makes preconditioning become expensive, or makes the treatment on obstacle constraints inflexible. The status quo of spatial clustering algorithm and obstacle constraints clustering algorithm both at home and abroad is studied in-depth in this paper. Spatial clustering algorithm and obstacle constraints clustering algorithm are compared and analyzed. For these deficiencies, taking mountain highway routing as the background an obstacle constraints-based density clustering algorithm with direction is presented in this paper. The clustering direction is introduced in to data space with obstacles. The algorithm only clusters the candidate points in the clustering direction, so that candidate points reduced. At the same time data partitioning is applied to partition the data set in the algorithm, the data are clustered in each district in order to avoid the sensitivity of initial parameter selection. Finally, the algorithm was experimentally verified and compared with DBCLuC algorithm. Experimental results show that the time complexity and clustering efficiency are greatly improved in the algorithm proposed in this paper. All these provide a good theoretical support for highway route.