Mining Obstacle Constraints On Space Co-location Model

Recent years, with the rapid development and widespread application of spatial information technology, a wide variety of spatial data show significant growth in quantity and complexity. However, the available spatial knowledge is deficient. Based on this, spatial association rule mining has an important significance for the development of various fields of space. Spatial co-location pattern mining is a special case of spatial association rule mining, which can find a group of spatial features whose instances are frequently associated in space. Up till now, there have been many achievements in co-location pattern mining field. However, these achievements are based on the idealized spatial data, ignoring the existence of constraints in the real world, such as obstacle constraints. In order to improve the practical value of the spatial co-location pattern mining, this paper investigates the spatial co-location pattern mining problem with obstacle constraints.Firstly, the paper introduces the basic concepts and algorithms of the spatial association rule mining, which illustrates the important significance of the spatial co-location pattern mining. Then we make a detailed introduction about the concepts, mining algorithms and current achievements. In addition, the paper introduces a current situation of spatial data mining with constraints and proposes the problems which the spatial co-location pattern mining with obstacles is confronting.Secondly, the paper proposes a solution on the co-location pattern mining with obstacle constraints. The feeble neighboring relation "FR" is suggested, based on which we give the related concepts:feeble co-location pattern、feeble table instance、 feeble participation index and so on. Then we propose an obstacle-oriented division algorithm which divides the collection of instances into standardized set of grids. We also propose a mining algorithm. In combination with the characteristics of instance distribution density in real world, we propose two pruning strategies to improve the efficiency of the algorithm. Thirdly, the algorithm is verified through experiments with simulation dataset. We analyze the impact of various parameters on the algorithms, and prove the correctness and validity of the algorithms. Meanwhile, experiments prove the accuracy and efficiency of pruning strategies by comparing the optimized algorithms with the basic algorithm.Finally, the conclusion and future work are presented. Furthermore, we point out the shortcomings of the research which we expect to improve in future.