论文标题
分析与致命道路崩溃相关的因素：一种机器学习方法
Analyzing Factors Associated with Fatal Road Crashes: A Machine Learning Approach
论文作者
论文摘要
道路交通损伤在全球范围内造成了巨大的人类和经济负担。了解导致致命伤害的风险因素至关重要。在这项研究中，我们提出了一个模型，该模型采用了由顺序最小优化和决策树构成的混合集合机学习分类器，以识别导致致命道路伤害的风险因素。该模型是使用黎巴嫩道路事故平台（LRAP）数据库构建，训练，测试和验证的，该数据库是8482道路崩溃事件，死亡人数是结果变量。进行了灵敏度分析，以检查多种因素对死亡发生的影响。在九个选定的自变量中，有7个与发生死亡的发生显着相关，即崩溃类型，伤害严重程度，空间群集ID和崩溃时间（小时）。从模型数据分析中获得的证据将由政策制定者和主要利益相关者采用，以了解与致命道路崩溃相关的主要因素，并将知识转化为安全计划和增强的道路政策。
Road traffic injury accounts for a substantial human and economic burden globally. Understanding risk factors contributing to fatal injuries is of paramount importance. In this study, we proposed a model that adopts a hybrid ensemble machine learning classifier structured from sequential minimal optimization and decision trees to identify risk factors contributing to fatal road injuries. The model was constructed, trained, tested, and validated using the Lebanese Road Accidents Platform (LRAP) database of 8482 road crash incidents, with fatality occurrence as the outcome variable. A sensitivity analysis was conducted to examine the influence of multiple factors on fatality occurrence. Seven out of the nine selected independent variables were significantly associated with fatality occurrence, namely, crash type, injury severity, spatial cluster-ID, and crash time (hour). Evidence gained from the model data analysis will be adopted by policymakers and key stakeholders to gain insights into major contributing factors associated with fatal road crashes and to translate knowledge into safety programs and enhanced road policies.
