基于平稳化长短期记忆网络模型的交叉口交通流量预测

Intersection traffic flow forecasting based on the smooth long short-term memory model

城市交叉口的道路通行能力目前已成为影响居民出行效率的重要因素之一。通过正向雷达道路车辆检测器采集过车数据,利用深度学习技术Tensorflow框架构建平稳化长短期记忆网络(long short-term memory,LSTM)交通量预测模型。该模型根据预测精度自适应机制更新层数和隐藏单元数,同时为了减小交通流序列的随机性,通过季节性时间差分方法对原始交通时间序列的输出特征序列进行了平稳化处理。在使用相同数据集条件下,将所提算法的预测结果与传统算法的结果进行对比,结果表明:平稳化的LSTM网络算法能有效提高交通流量预测的准确性,实际值与预测值拟合度超过95.5%,且在均方根误差和平均百分比误差上有较大的提升,该模型可为治理交通拥堵提供依据。

The road capacity of urban intersections has become one of the important factors affecting residents′ travel efficiency. Through the forward radar road vehicle detector to collect passing data, the data mining and analysis can gather the road condition information, so as to provide a basis for the management of traffic congestion. This paper uses the deep learning technology to build a smooth long short-term memory(LSTM) traffic volume prediction model to solve the problem of time series traffic volume prediction at intersections. The model uses the Tensorflow framework of in-depth learning to construct the hierarchical structure of LSTM network, and updates the number of layers and hidden units according to the prediction accuracy adaptive mechanism. At the same time, in order to reduce the randomness of traffic flow sequence, the output characteristic sequence of the original traffic time series is smoothed by seasonal time difference method. Under the same data set, the comparison with the prediction results of GRU model and support vector regression model shows that the smoothed LSTM network algorithm can effectively improve the accuracy of traffic flow prediction. The fitting degree between the actual value and the predicted value is more than95.5%. Moreover, the root mean square error and the average percentage error are greatly improved. This model can provide reference for easing the traffic congestion.