An Uncertain Continuous Nearest-Neighbor Query Based On The Conceptual Partitioning

The analysis based on uncertain data is one of the hot topics in the research of data mining and knowledge discovery due to its objectivity and reality. The traditional query treatment techniques can not play its role effectively because of the inherent magnanimity and complexity of the spatial data. Therefore, how to provide a variety of efficient space and spatial object query technologies is a problem needed to be solved currently. After putting forward the questions of nearest neighbor query on spatial data, whether in the field of spatial data mining theory, or a variety front-running practical application, studies related to the continuous development and extension, in order to solve the query that we face in our real life better. A variety of different index structures using spatial database query algorithms has been proposed, most of which are based on R-tree index structure of the nearest neighbor query algorithm. As the current query algorithms are the nearest neighbor queries under specified data, which is not effective to solve the nearest neighbor query on uncertain data, it is difficult to apply these query algorithms to the uncertain spatial data mining and related area. In the current study at home and abroad, the work involving continuous nearest neighbor query on uncertain object is limited, and have not formed a more complete theoretical system and without sophisticated algorithms supported.This thesis analyzes and summarizes the current study fruits and methods of the nearest neighbor query, based on uncertainty spatial data mining, with the help of a more efficient thought of the current concept partitioning using grid, and try to provide an index tree using the concept partitioning of the grids algorithm for continuous nearest neighbor query, through a series of steps to improve and enhance the efficiency of this method of inquiry. And on this basis, we extended this algorithm to the uncertainty of continuous nearest neighbor query problem on spatial objects, proposed a new ideas and methods which can solve the uncertainty spatial data in the continuous nearest neighbor query. And we have validated the correctness and validity of this method in tests.The main contribution of this paper can be summarized as follows:1)This paper proposed a method indexing conceptual partitioning grids using a tree structure.2)Designed a more efficient algorithm, which is a nearest neighbor queries, T-CPM algorithm, and which optimizes the order of grid search, and saves computational cost compared with the classical algorithms.3)We proposed an idea and method of continuous nearest neighbor query data based on concept partitioning.