多子群遗传神经网络模型用于路口短时交通流量预测

Multi-Population Genetic Neural Network Model for Short-term Traffic Flow Prediction at Intersections

为适应交叉路口短时交通流量的实时变化性和非线性性,提出将一种多子群遗传神经网络算法(MPGNN)应用于交叉路口短时交通流量的预测,结合BP网络对非线性问题良好的求解能力和遗传算法优良的全局寻优能力,建立遗传算法的多个子种群来搜索BP网络的最佳结构。通过对武汉市珞瑜路、武珞路、珞狮南路、珞狮北路交叉路口的短时交通流量进行预测分析,取得了良好的实验效果。根据预测结果对该路段的交通流量进行重新分配和控制,对缓解珞瑜路和武珞路段高峰时期交通流量的压力具有重要作用。

As the basis and a key issue of intelligent transportation system(ITS),researches on approaches for traffic flow prediction are significant.To adapt to the real-time characteristics and nonlinear feature of short-term traffic flow prediction at intersections,this paper presents a Multi-Population Genetic Neural Network(MPGNN)model for traffic flow prediction.MPGNN model combines the ability of BP neural network for solving nonlinear problem and the ability of genetic algorithm(GA)for global optimization.And...