改进SSA优化的BP神经网络交通量预测模型

为更加准确地进行交通量预测,针对传统的BP神经网络随机赋值、收敛速度慢等问题,提出了改进麻雀搜索算法（sparrow search algorithm,SSA）优化的BP神经网络预测模型。该模型结合SSA位置更新原理和鸡群优化算法中公鸡位置更新方法对麻雀搜索算法进行改进,在避免算法陷入局部最优和位置更新无效的同时有效地提高了算法的收敛速度。利用改进麻雀搜索算法对BP神经网络的权值和阈值进行寻优赋值,得到了改进SSA-BP神经网络预测模型。利用交通量数据,对LSTM神经网络、BP神经网络、SSA-BP神经网络和改进SSA-BP神经网络4种预测模型进行训练和测试,以MAE、MAPE、MSE、RMSE和EC 5个指标对预测结果进行对比分析。结果表明:BP神经网络优于LSTM神经网络,且麻雀搜索算法优化BP神经网络预测模型相较于BP神经网络预测模型MAE降低了0.28 veh/(3 min)、MAPE降低了1%、MSE降低了2.72 veh/(3 min)、RMSE降低了 0.04；改进麻雀搜索算法优化BP神经网络预测模型相较于BP神经网络预测模型MAE降低了1.31 veh/(3 min)、MAPE降低了4%、MSE降低了9.2 veh/(3 min)、RMSE降低了0.18,且拟合度更接近于1。改进SSA-BP预测模型的性能优于SSA-BP神经网络预测模型,且有效提高了BP神经网络的预测精度,拟合度达到0.98,该模型适用于交通量预测,能够为智能交通系统提供可靠的预测值。

Traffic volume forecast model based on BP neural network optimized by improved sparrow search algorithm

ln order to make traffic prediction more accurately, aiming at the problems of random assignment and slow convergence of the traditional BP neural network, an improved sparrow search algorithm(SSA) is proposed to optimize the BP neural network prediction model. The model combines the SSA position update principle and the rooster position update method in the chicken optimization algorithm to improve the sparrow search algorithm, which avoids the algorithm from falling into the local optimum and the ineffectiveness of the position update, and at the same time effectively improves the convergence speed of the algorithm. The improved sparrow search algorithm is used to optimize the weights and thresholds of the BP neural network, and the improved SSA-BP neural network prediction model is obtained. The four prediction models, namely, LSTM neural network, BP neural network, SSA-BP neural network and improved SSA-BP neural network, were trained and tested with traffic data, and the prediction results were compared and analyzed in terms of MAE, MAPE, MSE, RMSE and EC. The results show that the BP neural network is better than the LSTM neural network, and the optimized BP neural network prediction model with sparrow search algorithm reduced MAE by 0.28 veh/(3 min), MAPE by 1%, MSE by 2.72 veh/(3 min), and RMSE by 0.04 compared with the BP neural network prediction model; the optimized BP neural network prediction model with improved sparrow search algorithm reduced MAE by 1.31 veh/(3 min), MAPE by 4%, MSE by 9.2 veh/(3 min), RMSE by 0.18, and the goodness-of-fit was closer to 1. The improved SSA-BP prediction model outperforms the SSA-BP neural network prediction model and effectively improves the prediction accuracy of the BP neural network with a goodness-of-fit of 0.98, which is suitable for traffic volume prediction and can provide reliable prediction values for intelligent transportation systems.