中文摘要：

根据中国现行交通事故严重程度分类与事故信息数据分布特征,基于C5.0决策树方法,选取某省会城市城区及周边重点公路16 009起交通事故现场数据,分别将事故严重程度输出变量按照2分类和3分类,输入变量按照空间属性、涉事驾驶人及车辆属性和全属性,建立事故严重程度预测模型,生成相应规则集并利用测试样本进行检验和模型对比。研究结果表明：2分类和3分类事故严重程度预测模型精度分别为70%和61%,多模型综合优度有所提升;实证规则集揭示了影响事故严重程度分类的因素主要有,碰撞类型、道路属性、事故致因和驾驶人类型等。

英文摘要：

Based on the algorithm of C5.0decision tree,current severity classification of traffic accidents and the distribution characteristics of accident information data,this paper used the field data of 16,009 traffic accidents which occurred in some main highways of the urban area in and around a certain capital city to analyze the accident severity,and established a prediction model of accident severity according to the output variables based on dichotomy and trichotomy as well as the input variables based on the spatial attributes,the driver involved,vehicle attributes and the overall attributes.Through the test,this paper got the appropriate rule set and used the test samples for inspection and the comparison of models.The results show that the accuracy of the prediction model in accident severity is 70%and 61%separately based on dichotomy and trichotomy,and the integrated goodness of multi-model is improved.The empirical rule set reveals that the factors influencing accident severity classification are mainly the type of collision,road attributes,accident causation and the type of driver.5tabs,2figs,14 refs.