基于OD数据的群体行为可视分析  

Visual Analysis of Group Behavior Based on Origin-Destination Data

作　　者：黄文达 陶煜波[1] 屈珂 林海[1] Huang Wenda;Tao Yubo;Qu Ke;and Lin Hai(State Key Laboratory of CAD＆CG, Zhejiang University, Hangzhou 310058)

机构地区：[1]浙江大学CAD&CG国家重点实验室,杭州310058

出　　处：《计算机辅助设计与图形学学报》2018年第6期1023-1033,共11页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61472354,61672452); NSFC-广东省人民政府大数据科学研究中心项目(U1611263); 中央高校基本科研业务费

摘　　要：针对已有的公共自行车群体租车行为研究仅在站点尺度上进行分析,存在分析不完备等问题,提出了从城市—区域—站点的多尺度群体租车行为交互分析系统.首先,基于租/还站点和时间的一致性,从公共自行车数据中提取租车群体;然后结合站点的地理位置和群体租车行为的相似性,利用改进的迭代双聚类算法生成区域;再设计日历图、流量散点图和群体行为分布地图等视图,支持在宏观上分析群体行为随时间的变化和自流量与总流量的关系,在区域或站点尺度上分析群体行为的微观变化;最后结合多视图联动和交互分析,比较区域或站点的群体行为在工作日和周末的模式.通过4个案例,展示了群体租车行为的时空模式,证明了该系统在群体行为分析上的有用性和有效性.While previous methods usually analyze the group cycling behavior only at the station level, which results in incomplete analysis results, this paper proposes an interactive visual analysis system to investigate the group cycling behavior at multiple levels, from city to region to station. First, according to the consistency of location and time, groups are extracted from the public bicycle data. Then, the modified bi-clustering algorithm is used to generate regions based on the similarity of geographic location and cycling behavior. Next, multiple views, such as calendar view, scatter plot and group distribution map, are designed to analyze both the time variation of group behavior and the relationship between self-flow and total-flow at the macro level, and to explore micro changes on the region or station level. Finally, these coordinated views together with user interactions are combined to compare patterns of group behavior on weekend and weekday. Four case studies are conducted to analyze the spatial-temporal patterns on group behavior, and this demonstrates the effectiveness and usefulness of our system.

关 键 词：出发点-目的地数据 群体行为 区域划分 可视分析 时空数据 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]