城市交通事故多发点鉴别与时空模式挖掘  

Identification of black spots in urban roads and spatiotemporal patterns mining

作　　者：吴佩洁 孟祥海[1] 曹梦迪 WU Peijie;MENG Xianghai;CAO Mengdi(School of Transportation Science and Engineering,Harbin Institute of Technology,Harbin Heilongjiang 150090,China)

机构地区：[1]哈尔滨工业大学交通科学与工程学院,黑龙江哈尔滨150090

出　　处：《中国安全科学学报》2020年第11期127-133,共7页China Safety Science Journal

基　　金：国家自然科学基金资助(71701055)。

摘　　要：为准确、快速、微观地识别城市路网中交通事故多发点及其时空演变特性,降低交通事故发生频率,提出一种基于时空立方体的事故数据挖掘方法。首先,探究构建时空立方体模型时重要参数选择方法,通过时空立方体和累计频率曲线法,识别出城市路网中的事故时空多发点;然后,基于新兴时空热点分析法,研究事故多发点的时空动态演变模式;最后,对比分析时空立方体法与质量控制法、经验贝叶斯法和核密度估计法的优缺点及适用范围。研究结果表明:时空立方体法和核密度估计法适合基于小样本事故数据的中微观尺度下的事故多发点鉴别,可采用新兴时空热点分析法得到事故多发点的不同时空演变模式。In order to identify black spots and their spatiotemporal evolution patterns in urban road networks accurately,rapidly and microcosmically,and to effectively reduce frequency of traffic accidents,a crash data mining method based on space-time cubes was proposed.Firstly,selection method of important parameters in construction of space-time cubes model was studied,and spatiotemporal black spots in urban road network were identified by"space-time cubes+cumulative frequency curve method".Then,their spatiotemporal dynamic evolution patterns were obtained based on emerging hot spot analysis method.Finally,the space-time cubes method was compared with quality control method,empirical Bayesian method and kernel density estimation from aspects of their advantages,disadvantages and application scopes.The results show that space-time cubes method and kernel density estimation are applicable to identification of black spots at the meso and micro scale while emerging hot spot analysis method can be utilized to obtain their spatiotemporal evolution patterns.

关 键 词：城市交通事故 事故多发点 时空模式 时空立方体 中微观尺度 

分 类 号：X928.02[环境科学与工程—安全科学]