面向交通状态时空模式的双向聚类可视分析

Visual Analysis of Temporal and Spatial Patterns of Traffic State Based on Co-clustering

城市道路网络的交通状态存在多种时序变化模式和空间分布模式.为了能综合分析各时序模式在不同区域的分布情况,以及各空间模式在不同时间段的出现规律,本文提出一种基于双向聚类的时空模式可视分析方法.该方法可同时将路段集合和以小时为尺度的时间戳集合划分为簇,以提取出时序模式和空间模式.然后,通过多视图联动和降维投影可视化分别从小时尺度和日期尺度上分析时序模式的空间分布,同时分析空间模式的时间分布.本文以一份新冠疫情期间采集的交通数据为例进行研究,实验结果表明,该方法能有效发现因政府交通限制等因素而呈现的多种时序模式和空间模式,同时辅助用户分析这些模式在时空上的分布情况.

The traffic state of the urban road network has a variety of temporal change patterns and spatial distribution patterns.In order to comprehensively analyze the distribution of various time series patterns in different regions and the appearance of each spatial pattern in different time periods, this paper proposes a visual analysis method of spatiotemporal patterns based on two-way clustering.This method can divide the road sections and the hourly timestamps into clusters at the same time to extract the time series pattern and the space pattern.Then, through multi-view linkage and dimensionality reduction projection visualization, the spatial distribution of time series patterns are analyzed on the hour scale and date scale, and the time distribution of spatial patterns is analyzed at the same time.This article uses a traffic data collected during the COVID-19 pandemic as an example to study.The experimental results show that this method can effectively find a variety of time series patterns and spatial patterns caused by government traffic restrictions or other factors, and assist users in analyzing the distribution of these patterns in time and space at the same time.