基于多尺度空间划分与路网建模的城市移动轨迹模式挖掘

针对城市移动轨迹模式挖掘问题展开研究, 提出移动全局模式与移动过程模式相结合的挖掘方法, 即通过移动轨迹的起始位置点--终点位置点 (Origin-destination, OD点) 与移动过程序列分别进行移动全局模式与过程模式的发现. 在移动全局模式发现中, 提出了弹性多尺度空间划分方法, 避免了硬性等尺度网格划分对密集区域边缘的破坏, 同时增强了密集区域与稀疏区域的区分能力.在移动过程模式发现中, 提出了基于移动轨迹的路网拓扑关系模型构建方法, 通过路网关键位置点的探测抽取拓扑关系模型.最后基于空间划分集合与路网拓扑模型对原始 移动轨迹数据进行序列数据转换与频繁模式挖掘. 通过深圳市出租车历史 GPS 轨迹数据的实验结果表明, 该方法与现有方法相比在区域划分、数据转换等方面具有更好的性能, 同时挖掘结果语义更为丰富, 可解释性更强.

Mining Urban Moving Tra jectory Patterns Based on Multi-scale Space Partition and Road Network Modeling

In this paper, the problem of discovering moving trajectory patterns in urban environment is studied and the method of integration of moving global pattern and moving local pattern is proposed. Through moving trajectory origin-destination (OD) and moving sequence features, the global patterns and local patterns are mined. In the process of moving global pattern mining, a flexible multi-scale space partition is devised to avoid damage of the dense region edges by hard regular grid division and enhance the ability to distinguish the dense regions and sparse regions. In the process of moving local pattern mining, the modeling method of road network based on moving trajectory is devised to extract the feature of topological relation by key road network nodes. Finally, the raw moving trajectory dataset is converted by partitioned discrete regions and road network model, and the frequent moving trajectory patterns are discovered by a modified sequence pattern mining algorithm. A comprehensive experimental evaluation on Shenzhen taxicabs GPS trajectory dataset is presented, and the evaluation shows that the proposed method outperforms the existing methods in space division, data transform, and interpretability of mined patterns.