基于PSO-LSTM的高速公路短时交通流预测

Short-term Highway Traffic Flow Prediction Based on PSO-LSTM

为了实现高速公路短时交通流的精准预测,提出了考虑天气因素的PSO-LSTM深度学习预测模型以提高预测准确性。首先,对获取到的高速公路交通流数据、天气数据进行异常值处理、数据去噪、数据归一化等预处理;其次,构建预测模型,并采用粒子群优化算法寻找LSTM模型的最优参数组合,需要寻优的参数有神经网络最大迭代次数、学习率以及隐藏层神经单元个数;最后,通过PEMS网站提供的交通流数据进行实例验证,将交通流数据分为工作日与非工作日。结果表明,本文提出的预测模型能够较好的描述高速公路交通流变化规律,且相比于BP、SVM以及不考虑天气因素的PSO-LSTM模型,具有较高的预测精度。

In order to achieve accurate prediction of short-term traffic flow on expressway,a PSO-LSTM deep learning prediction model considering weather factors is proposed to improve the prediction accuracy.Firstly,the obtained highway traffc flow data and weather data are preprocessed by outlier processing,data denoising,data normalization,etc.;Secondly,the prediction model is constructed,and the particle swarm optimization algorithm is used to find the optimal parameter combination of the LSTM model.The parameters to be optimized include the maximum number of iterations of the neural network,the learning rate,and the number of hidden layer neural units;Finally,the traffic flow data provided by PEMS website is used for example verification,and the traffic flow data is divided into working days and non working days.The results show that the prediction model proposed in this paper can better describe the change law of highway traffic flow,and has higher prediction accuracy compared with BP,SVM and PSO-LSTM models without considering weather factors.