ITR: Advanced Algorithms for Spatial-Temporal Interactions in Distributed GIS

本研究的重点是分布式GIS环境下的空间交互分析。与空间相互作用系统有关的数据包括空间流动以及与起源和目的地有关的位置和属性数据。鉴于这些数据集分布在计算机网络的多个节点上，本研究旨在：1)研究数据分区的机制，并开发描述分区的元数据结构；2)开发可分解的重力建模算法，使通信成本最小化；3)利用分布式数据源开发高效的流模式学习和分类算法。方法是让数据库驻留在它们的本机站点上。这些算法将自己动态分解为在各个数据库站点上执行的部分计算，并且组合本地结果以获得与将数据库移动到公共站点时所获得的相同的全局结果。该算法可以找到分解来匹配数据在整个网络中的分布。本研究将对许多需要处理分布式数据的问题产生重大影响。例如，它可以使GIS系统分析分布在互联网上的时空数据集，而不必将数据库移动到一个公共站点。

This research focuses on the analysis of spatial interactions in distributed GIS environments. Data related to systems of spatial interactions includes spatial flows and locational and attribute data pertaining to origins and destinations. Given that these datasets are distributed across multiple nodes of a computer network, this research aims to: 1) study mechanisms of data partitioning and develop a metadata structure that describes the partition; 2) develop decomposable algorithms for gravity modeling that minimize communication cost; and 3) develop efficient algorithms for learning and classifying flow patterns using distributed data sources. The approach is to let the databases reside at their native sites. The algorithms dynamically decompose themselves into partial computations that are executed at individual database sites, and local results are composed to obtain the same global results that would have been obtained if the databases had been moved to a common site. The algorithms can find the decompositions to match the distribution of data across the network. This research will have significant impact on many problems that need to process distributed data. For example, it can enable GIS systems to analyze spatial-temporal datasets distributed over the Internet without having to move the databases to a common site.