基于模糊神经网络的交通事故微观预测方法研究

Research on Microcosmic Forecast Methods of Road Accidents Based on Fuzzy Neural Network

道路交通事故微观预测包括对路段和交叉口事故指标的预测。本文总结现有预测方法的优劣性,探讨现有预测方法的改善方向,提出了基于模糊神经网络的交通事故微观预测方法,分析了网络结构和学习算法。以石河子市交通事故调查数据进行实例分析,选择路段事故影响因素作为输入变量,通过Matlab编程实现模糊神经网络的算法,并与负二项回归模型、BP神经网络模型作出比较,证明了模糊神经网络模型的优越性。

The microcosmic prediction of road accidents consists of the prediction of highway section and intersection accidents. Based on an overall summary about the existed superior and inferior forecast methods, the improvement direction in the existed prediction methods was discussed, an accident microcosmic prediction method based on fuzzy neural network wasput forward, then the network structure and learning algorithm were analyzed; an instance analysis was carried out with the traffic accident census data of Shihezi. An influence factor of highway section was selected as an import variable, then, the algorithm of fuzzy neural network was realized through Matlab program. Compared with the negative binomial regression model and BP neural network, the advantage of the fuzzy neural network model was examined.