高速公路事故多发点鉴别及诱发因素识别

为了提高事故多发点鉴别的客观公正性及自动化识别水平，提出了在公路沿线上划分初始评估地点的动态聚类算法，建立了鉴别事故多发点的自组织神经网络模型，给出了基于离散多变量算法与概率分布相结合的事故多发点突出事故诱发因素识别过程. 方法应用结果表明，基于动态聚类的初始评估点划分方法能够客观地描述出事故点在公路沿线上的集中与分散状况，而神经网络鉴别模型能够对初始评估地点的安全状况进行自动分类且结果较合理. 在掌握了能够满足统计分析要求的事故样本点数量的基础上，能够应用突出事故诱发因素识别方法建立一套评估标准，并用来识别事故多发点的突出事故诱发因素.

Research on Identification of Black Spot and Accident Inducing Factor for Freeway

In order to improve the objectivity, fairness and automated identification level of black spot identification, dynamic clustering algorithm, which is used to divide initial assessment of sites along the road, is proposed firstly. Then, the self-organizing neural network model is established to identify black spot. Thirdly, black spot prominent accidents induced factors’ identification process based on discrete multi-variable algorithm combined with probability distribution is presented. The results show that the initial assessment based on dynamic clustering method can describe the accidents’ concentration and dispersion objectivity, while the neural network model can give an automatic classification to the initial assessment of the security situation and the results are more reasonable. In the master can meet the requirements of statistical analysis of accidents on the basis of the number of sample points, black spot prominent accidents inducing factor identification methods, which can be used to establish a set of evaluation criteria and identify prominent accidents inducing factor.