基于Bagging算法和遗传神经网络的交通事件检测

TRAFFIC INCIDENT DETECTION BASED ON BAGGING METHOD AND GENETIC NEURAL NETWORK

提出一种集成遗传神经网络的交通事件检测方法,以上下游的流量和占有率作为特征,RBF神经网络作为分类器进行交通事件的自动分类与检测。在RBF神经网络的训练过程中,采用遗传算法GA(Genetic Algorithm)对RBF神经网络的隐层中心值和宽度进行优化,用递推最小二乘法训练隐层和输出层之间的权值。为了提高神经网络的分类能力,采用Bagging算法,进行网络集成。通过Matlab仿真实验,证明该方法相对于传统的事件检测算法能更准确、快速地实现分类。

A method based on integrated RBF neural network is proposed for traffic incidents detection.Taking the upstream and downstream flows and occupancy rate as the features,RBF neural network is used as a classifier to automatically classify and detect the traffic incidents.The genetic algorithm is used to optimize the hidden layer centre＇s value and width of RBF neural network and the recursive least square method is used to train the weights between hidden layer and output layer during the training of the RBF neural network.In order to improve the classifying ability of the RBF neural network,Bagging algorithm is used to build an integration neural network.Through the simulation experiment on Matlab we found out that the algorithm can achieve a more accurate and fast classification than traditional incident detection algorithms.