Research On Spatial Association Pattern Mining And Updating Based On Distributed Computing

Spatial association pattern mining is a process of mining the knowledge related to the location from the spatial database. Spatial association pattern mining from large scale spatial data is one of the methods to rapidly transform spatial data into knowledge. With the development of spatial information technology, people have more and more spatial data. There are two main problems in the association pattern mining from massive spatial data. Firstly, the stock of spatial data is large, the traditional single machine spatial association pattern mining algorithm is unable to complete the task of mining. Secondly the updating of spatial data is frequent. According to the traditional way, each update will need to scan all the spatial data to update the spatial association pattern. Since the new data is only a small part of the original data, it is a great waste to mining the spatial association patterns with all data.In view of the problems of large amount of spatial data and frequent updating of spatial data, this paper proposes a distributed spatial association pattern mining algorithm based on incremental updating. The algorithm of mining spatial association patterns from the total amount data and the algorithm of updating spatial association patterns with the incremental data are executed alternately. In order to adapt to the distributed computing, this algorithm first uses Hilbert space filling curve code of the spatial object for data partition firstly. In this paper, we propose an algorithm library that is the collection of spatial relation analysis algorithms for computing of distributed spatial transaction set. On the basis of the rational distribution of spatial transactions, the spatial association pattern mining algorithm is used to search spatial association patterns. In order to update the spatial association patterns rapidly when new spatial data is added, the frequent spatial association patterns and the boundary spatial association patterns are preserved in the process of spatial association pattern mining from the total amount data. The boundary spatial association patterns is used to buffer the change of frequent spatial association patterns brought about by the newly added spatial data.Distributed mining and updating of spatial association patterns can greatly improve the computational efficiency, and can provide users with the latest spatial knowledge in real time.Finally, the experimental results show that the efficiency of algorithm in this paper is better than traditional algorithm.