Visual Analysis of Group Behavior Based on Origin-Destination Data

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.