基于改进灰狼算法优化LSTM的断面交通流预测

LSTM for Cross-sectional Traffic Flow Prediction Based on Improved Grey Wolf Optimizer Algorithm

为提高断面交通流的预测性能,提出一种基于改进灰狼优化算法(IGWO)优化长短期记忆神经网络(LSTM)的断面交通流预测方法。首先,针对灰狼优化算法(GWO)中存在的狼群易陷入局部最优等问题提出了非线性收敛因子、基于SPM混沌映射的种群初始化和基于权重的位置更新策略3个方面的改进。其次,建立IGWO优化LSTM网络超参数的断面交通流预测模型(IGWO-LSTM),同时构建并行子模型。最后,将IGWO-LSTM应用于断面交通流预测,结果表明所提出的IGWO-LSTM模型在断面交通流中具有更好的预测性能。此外,还研究了所提出的模型在跨步断面交通流中的预测效果,以测试和验证所提出的模型。

In order to improve the prediction performance of cross-sectional traffic flow,aprediction method based on the improved grey wolf optimization(IGWO)algorithm for optimizing long-short term memory(LSTM)neural network is proposed.Firstly,three improvements in terms of nonlinear convergence factor,population initialization based on SPM chaotic mapping and weight-based location update strategy are proposed to address the problems of the grey wolf optimization(GWO)algorithm that wolves are prone to fall into local optimum.Secondly,the IGWO optimized LSTM network’s hyper-parameters for cross-sectional traffic flow prediction model(IGWO-LSTM)is established,while parallel submodels are constructed.Finally,IGWO-LSTM is applied to the cross-sectional traffic flow prediction.And the results show that the proposed IGWO-LSTM model has better prediction performance in cross-sectional traffic flow.In addition,this paper also investigates the effect of the proposed model in cross-step cross-sectional traffic flow prediction to test and validate the proposed model.