基于概率神经网络的高速公路交通事故多发点安全预警模型

针对高速公路交通事故多发点交通事故难以预测的问题,利用神经网络的非线性逼近能力,结合概率神经网络(PNN)模式分类功能建立安全预警模型。设计概率神经网络拓扑结构,给出交通状态模式类别,确定相应交通事故指标体系,概述概率神经网络的学习过程,并通过Matlab仿真实验对其性能进行了测试。结果表明:采用PNN神经网络辨识技术的网络模型预警准确率高、泛化能力强,可对高速公路交通安全进行实时监测,对有效预防和控制交通灾害的发生是完全可行的。

An Early-Warning Model of Freeway Hazardous Locations Based on Probabilistic Neural Network

Traffic accidents on freeway hazardous locations are hard to predict, to solving this problem, an early-warning model was made by using the nonlinear approximation capability with the pattern classification function of the probabilistic neural network （PNN）. By designed the probabilistic neural network topology structure, provided traffic state categories, determined the index system of related traffic accidents, sketched out the learning process of the probabilistic neural network, the properties were also tested via the Matlab simulation experiment. Results indicate that, the early-warning model with the PNN recognition technology achieves quite high detection accuracy, and the ability of generalization is well, can be used at freeway traffic safety real-time monitoring, and as an effective prevention and control approach against the factors causing road traffic hazards is entirelv possible.