改进ACO优化的BP神经网络短时交通流量预测

Improved ACO-Optimized BP Neural Network For Short-Term Traffic Flow Prediction

针对BP神经网络算法收敛速度缓慢、易陷入局部最小值、在短时交通流量预测的问题中精度不高等问题,提出了一种改进ACO(蚁群算法)优化的BP神经网络短时交通流量预测算法。在确定BP神经网络权阈值的过程中,采用蚁群信息素挥发自适应参数、在蚁群信息素更新时采用精英选择策略和种群更新时加入变异因子的方法来得到最优权阈值。仿真结果表明,改进算法在预测流量趋势和准确度方面均有较大提升,在短时交通流量预测方面取得了良好的效果。

Aiming at the problems of slow convergence speed, easy to fall into local minimum and low accuracy in short-term traffic flow prediction of BP neural network algorithm, this paper proposes an improved ACO(ant colony algorithm) optimized BP neural network prediction algorithm for short-term traffic flow. In the process of determining BP neural network weight threshold, the optimal weight thresholds were obtained by using the ant colony pheromone volatilization adaptive parameter, the elite selection strategy when the ant colony pheromone was updated, and the variation factor when the population was updated. Simulation results show that the proposed algorithm has significant improvement in forecasting traffic trends and accuracy, and the algorithm has achieved good results in short-term traffic flow prediction.