基于深度学习的车流量预测方法研究

Research on Traffic Flow Prediction Method Based on Deep Learning

交通流量信息在智能交通系统管理中起着重要作用。随着交通流检测、采集技术的发展,海量的交通信息得以获取,为提高预测准确性提供了机会。论文建立了基于自动编码器和LSTM递归神经网络的交通流量预测模型,该模型首先利用自动编码器进行无监督的特征表示学习,训练自动编码器层参数,然后将自动编码器隐含层输出作为LSTM层输入,利用期望输出和实际输出误差调整LSTM层和输出层参数,分析挖掘北京市朝阳区的路口交通数据,实验结果表明提出的预测模型能够更好地反映路口交通流的变化特征。

Traffic flow information is very important in the management of intelligent transportantion system. With the develop. ment of traffic flow detection and acquisition technology,massive traffic information is acquired,which provides an opportunity to improve the accuracy of forecasting. In this paper,a traffic flow forecasting model based on autoencoder and LSTM recurrent neural network is established. The model first uses an autoencoder for unsupervised feature representation learning and trains the autoen. coder layer parameters. Then the output of the autoencoder's hidden layer is the input to the LSTM layer. The LSTM layer's parame. ters and output layer's parameters are adjusted by the expected output and actual output error. We analyze the traffic data at the inter. section of Chaoyang District in Beijing. The experimental results show that the model can better reflect the changing characteristics of intersection traffic flow.