基于粒子群优化的神经网络短时交通流量预测

Neural network short-term traffic flow forecasting model based on particle swarm optimization

基于短时交通量的不确定特性,对城市相邻交叉口路段的交通流建模方法进行了研究。提出了基于粒子群优化的BP神经网络的信号交叉口交通量预测模型。该模型以BP神经网络为基础,用PSO算法对BP神经网络权值和阈值进行优化,从而提高了网络的预测精度。实时预测时,不只考虑本路口前几个时段交通量,同时也考虑了上下游路段的交通量的影响。仿真结果表明,粒子群-BP神经网络预测模型可以成为交通量预测的一种有效手段。

Based on the indeterminacy of the short-term traffic flow, a study of the modeling method of the conjoint crossings are made. Aparticle swarm optimization（PSO）based BP neural network mode is developed to predict the traffic flows. Basedona BPNN, function and weights are optimized by PSO algorithm. Therefore it has enhanced forecasting accurate. In real-time prediction, it not only took into account several pre-period traffic flow of this junction, but also the up-down junctions＇ traffic flow. By the emulation the practice proves the PSO-neural network model is rather effective and it is used as a good prediction model.