多元地理流的空间丛集模式挖掘

地理系统中物质、能量、信息等在不同空间位置之间的移动、交换或关联形成地理流，如人流、物流、资金流等，常被抽象为流动起点（O）到终点（D）的OD流（简称流）。由于地理系统的复杂性，从流的视角探究地理现象背后的机理时，通常需要同时关注存在相互作用的多种流，本项目将其定义为多元流。对于地理流而言，空间模式是其最基本特征。因此，挖掘并解析多元流的空间模式是认识地理系统格局和演化的有效途径。然而，现有研究在挖掘多元流的空间模式时，存在模式分类不明确、模型参数选择主观性强等问题。为此，项目拟以多元流的空间丛集模式为例，将点过程的空间统计和空间模式挖掘方法拓展到流，试图构建：1）多元流的空间模式分类方案，2）多元流的空间丛集模式判别及定量化的无参数方法，3）多元流的空间丛集尺度定量化及丛集模式提取方法，从而实现以空间模式显著统计特征为依据的多元流模式挖掘，最终形成多元流空间模式挖掘的方法框架。

A geographical flow depicts the movements, exchanges or association of the materials, energy and information, etc., among the spatial locations in the geographical system. The examples of geographical flows are the human flow, the commodity flow and the capital flow, etc. In geographical research, a geographical flow is often abstracted as an OD flow (flow for short) which records the origin and destination of the flow. Due to the complexity of the geographical system, when exploring the mechanisms of the geographical phenomenon, various flows with interactions, which we defined as the multivariate flows, must be focused at the same time. For geographical flows, the spatial patterns are the most basic characteristics. Therefore, Mining and analyzing the spatial patterns of multivariate flows is an effective mean for the cognition of the patterns and evolutions of the geographical system. However, studies of the pattern mining of multivariate flows have not made a specific classification for multivariate flows and the parameters of the models are subjective. To solve these problems, this project takes the cluster pattern as an example, and attempts build 1) the spatial pattern classification scheme, 2) the parameterless methods for spatial cluster pattern mining and quantification, and 3) the spatial cluster scale quantification and spatial cluster pattern mining methods of the multivariate flows, by extending the spatial statistics and spatial pattern mining methods of points to flows. From these progresses, this project intends to realize the spatial pattern mining of the multivariate flows based on the statistically significant characteristics, and eventually build a spatial pattern mining framework of the multivariate flows.