基于历史事故数据的事故预测模型是计量评价道路交通安全的主流方法。现有高速公路事故预测模型缺乏系统性的针对不同事故伤亡程度和事故类型的多元建模方法，鲜有考虑相邻路段安全因素对路段安全水平的交互影响（即空间溢出效应），以及从模型预测方面对事故漏报问题的数据修补技术。.本项目以深入探究高速公路事故发生机理、科学评价高速公路交通安全水平为目标，拟将事故频次按照伤亡程度、事故类型等多重指标进行划分，探索自身路段及相邻路段的道路、交通、环境属性与特定类别事故频次间的内在关联；在新兴的多元时空模型基础上，融入对空间溢出效应和事故漏报特性的考量，实现对高速公路事故风险、伤亡程度和事故类型的综合评价。应用层面上，开发高速公路黑点识别方法、安全改进措施和规划方案预期效应评估技术，为高速公路交通安全管理、设计和规划方案的制定提供技术支持和实践指导。

Crash prediction models based on historical crash data, are the prevalent methods for the quantitative evaluation on roadway safety. However, it is short of systematic modeling crash frequencies by injury severity and crash type in extant freeway crash prediction models. The interactive effects of the observed factors between adjacent freeway segments, are rarely investigated. Moreover, the crash underreporting issue is seldom taken into account in crash frequency modeling..To deeply explore the occurrence mechanism of freeway crashes and to accurately evaluate the safety performance of freeways, this project will investigate the underlying relationship between crash frequency by injury severity & crash type and the roadway, traffic, and environmental attributes of themselves and adjacent roadway segments. A novel crash prediction model will be proposed, on the basis of emerging multivariate spatio-temporal modeling technique and with consideration of spatial spillover effects and crash underreporting characteristic, for the comprehensive evaluation on freeway crash risk, injury severity and crash type. Based on the proposed crash prediction model, the methods for hotspot ranking and safety effects estimation on candidate countermeasures or planning schedules will be designed. The methods and tools are expected to provide new technical supports for prioritizing safety management, design, or planning projects on freeways in practice.