基于BMO算法混合神经网络的短时交通流预测

Short-Term Traffic Flow Prediction Based on BMO Algorithm and Hybrid Neural Network

准确预测短时间内某路段的交通流量,可以极大提升城市交通效率,而城市交通流预测的核心是各种交差路口附近的车流预测,尤以十字路口最为常见和复杂。针对具有极强的时空相关性且稳定性交差的情况,提出使用改进鸟类繁殖算法(Bird Mating Optimizer,BMO)混合BP神经网络(Back Propagation Neural Network,BPNN)模型对交通流进行非线性拟合。文章使用基于适应度方差的参数自适应调整策略改进了BMO算法,并结合模拟退火思路改善算法早熟问题。使用改进的BMO算法解决了训练时间长和收敛速度慢的缺陷。仿真结果显示,该模型具有更好的非线性拟合能力,使十字路口交通流预测准确率提高了11.4%。

Accurate prediction of the traffic flow of a certain section in a short time can greatly improve the urban traffic efficiency.The core of urban traffic flow prediction is the traffic flow prediction near the intersection with various traffic differences,especially the intersection is the most common and complex.In view of its strong spatiotemporal correlation and stable intersection,this paper proposes to use the improved Bird Breeding Optimizer(BMO)hybrid Back Propagation Neural Network(BPNN)model for nonlinear fitting of traffic flow.In this paper,the BMO algorithm is improved by using the adaptive parameter adjustment strategy based on fitness variance,and combined with the idea of simulated annealing to improve the premature problem of the algorithm.The improved BMO algorithm solves the defects of long training time and slow convergence speed.The simulation results show that the model has better nonlinear fitting ability,which improves the accuracy of traffic flow prediction at intersections by 11.4%.