基于时空热点分析的城市交通违法行为特征识别方法

交通违法行为是引发交通事故的重要原因,然而现有研究主要关注交通违法行为的整体特征,缺少面向交通违法治理需求的分析框架。本文基于时空热点分析方法,提出从热点区域时间分布特征和典型时段热点区域空间分布特征两个角度识别交通违法行为特征的分析框架,分别用于支撑局部交通违法热点以及全局违法模式的原因分析和精准治理。基于该方法对福州市的机动车和非机动车（含行人）违法行为特征进行了识别分析,结果表明：机动车和非机动车违法行为在时间维度均呈现出9:00和16:00一日双峰特征,在空间维度呈现出“一片区、多热点”的聚集分布特征。二者也存在明显差异,具体表现为：① 在时间维度,非机动车违法行为呈现出更大的变化幅度,高峰时段与中午低谷时段、工作日与周末的违法行为数量差异均明显高于机动车;② 在空间维度,机动车违法行为在商业中心、医院等重要场所和交通枢纽呈现出聚集特征,分布范围更广,而非机动车违法行为则主要在人流量大且人车混行严重的城市中心路口区域呈现聚集特征;③ 不同违法热点地区产生的原因存在差异,需要有针对性制定治理措施。上述发现表明了本文方法能够全面快速识别交通违法行为特征,可以帮助指导城市交通违法行为动态监测分析系统建设,为持续优化城市交通现场执法警力动态分配以及交通违法行为精准治理提供决策支持。

Study on the Method of Identifying the Characteristics of the Traffic Violation Behavior based on the Spatial and Temporal Hotspot Analysis Approach

Urban traffic violation behavior plays an important role in traffic accidents. Analyzing the spatial and temporal distribution of traffic violation behavior can support related decision makings for traffic management and the optimization of the surroundings of the hotspots. Due to the limitation in data acquisition, existing studies paid little attention to the variation of the spatial and temporal patterns between different violation behavior types. There is a lack of analysis framework to support the decision makings in traffic violation behavior treatment. In this study, we propose a traffic-violation-behavior treatment-oriented analysis framework based on the spatial and temporal hotspot approach. Two analyses are designed and conducted to support the traffic violation behavior treatment: (1) analyzing the temporal pattern of each spatial hotspot to support the reasoning analysis and precise treatment policy makings at the local scale; (2) analyzing the spatial pattern of the hotspots during typical periods (e.g., morning and evening rush hours) to support the reasoning analysis and the optimization of the allocation of police resources on a global scale. We use a dataset of Fuzhou city acquired in 2017 to verify the proposed method. The spatial and temporal patterns of the motor traffic violation behavior and the non-motor type are analyzed and compared. We find that: (1) the traffic violation behavior exhibits a double peak hourly pattern at 9:00 am and 4:00 pm during a day, respectively. The morning peak is obviously higher than the evening peak. The traffic violation behavior more likely happens during weekdays than weekends; (2) the traffic violation behavior mainly concentrates at the core-built area within the second ring highway and several hotspots in the suburban area including the shopping mall of Cangshan Wanda and the exit of the Kuiqi tunnel oriented to Mawei; (3) motor and non-motor traffic violation exhibit different temporal and spatial patterns. Non-motor traffic violation frequencies exhibit both larger hourly and weekday-weekend differences, and mainly concentrates at the road crosses with big traffic volume of both motor cars and e-bikes/pedestrian. While the motor traffic violation exhibits more stable patterns across the hours in a day and the days in a week, and mainly happens around the critical places such as large hospitals, shopping malls, and complex overpasses; (4) the spatial scales affect the patterns of the spatial hotspots of the traffic violation behavior. The spatial autocorrelation of the traffic violation increases with the scale size rapidly before 1500 m and keeps around 0.6 afterward. Motor traffic violation exhibits lower spatial autocorrelation than the non-motor. The above findings validate the effectiveness of the proposed method. It can help to guide the construction of the traffic violation behavior treatment platform and further optimize the allocation of the police resources and improve the effectiveness of the law enforcement for the traffic violation behavior.