针对现有预测模型不能充分提取交通流时空特征的问题，提出一种基于改进卷积神经网络（convolutional neural network，CNN）和长短时记忆（long short-term memory，LSTM）神经网络的短时交通流预测方法。首先，采用分层提取方法使设计的网络结构和一维卷积核函数自动提取交通流序列的空间特征；其次，优化LSTM网络模块来减少网络对数据的长时间依赖；最后，在端对端模型的训练过程中，引入改进后的自适应矩估计（rectified adaptive moment estimation，RAdam）优化算法，加快权重的拟合并提高网络输出的准确性和鲁棒性。实验结果表明：在工作日与周末分时段，所提出的模型相比堆栈自编码（stacked auto-encoders，SAEs）网络预测模型，性能分别提升3.55%与8.82%，运行时间分别缩减6.2%与6.9%；相比长短时记忆网络-支持向量回归（long-short term memory-support vector regression，LSTM-SVR）预测模型，性能分别提升0.29%与1.79%，运行时间分别缩减9.0%与9.7%。所提模型能够更加适用于不同时段下的短时交通流预测。
Aiming at solving the problem that existing prediction models could not fully extract the spatio-temporal features in traffic flow, we proposed an improved convolutional neural network (CNN) with long short-term memory neural network (LSTM) for shortterm traffic flow prediction. First of all, a layered extraction method was used to design the network structure and one-dimensional convolution kernel which enabled automatic extraction of spatial features of traffic flow sequences. Second, the LSTM network modules were optimized to reduce the long-term dependence of network on the data. Finally, the optimization algorithm for rectified adaptive moment estimation (RAdam) was introduced to the end-to-end model training process, which accelerated fitting effects of the weight and improved the accuracy and robustness of network output. Experimental results showed that compared with the prediction model of stacked auto-encoders (SAEs) network, performance of the proposed model was enhanced by 3.55% and 8.82% on weekdays and weekends with model running times reduced by 6.2% and 6.9%, respectively. Compared with the prediction model of long-short term memory-support vector regression (LSTM-SVR), its performance was enhanced by 0.29% and 1.79% with model running times reduced by 9.0% and 9.7%, respectively. Therefore, the proposed model was more applicable to the short-term traffic flow prediction of different time periods.
[1] Morocho M, Lee H, Lim W. Machine learning for 5G/B5G mobile and wireless communications:potential, limitations, and future directions[J]. IEEE Access, 2019, 7(99):137184-137206.
[2] Zhao J H, Ni S J, Yang L, et al. Multiband cooperation for 5G HetNets:a promising network paradigm[J]. IEEE Vehicular Technology Magazine, 2019, 14(4):85-93.
[3] Ma X T, Zhao J H, Gong Y, et al. Carrier sense multiple access with collision avoidance-aware connectivity quality of downlink broadcast in vehicular relay networks[J]. IET Microwaves, Antennas & Propagation, 2019, 13(8):13-19.
[4] 许伦辉, 唐德华, 邹娜, 等. 基于非线性时间序列分析的短时交通流特性分析[J]. 重庆交通大学学报(自然科学版), 2010, 29(1):110-113. Xu L H, Tang D H, Zou N, et al. Analysis on short-term traffic flow characteristics based on nonlinear time series analysis[J]. Journal of Chongqing Jiaotong University (Natural Science), 2010, 29(1):110-113. (in Chinese)
[5] Kumar S V, Vanajakshi L. Short-term traffic flow prediction using seasonal ARIMA model with limited input data[J]. European Transport Research Review, 2015, 7(3):1-9.
[6] 谈苗苗, 成孝刚, 周凯, 等. 基于ARIMA和灰色模型加权组合的短期交通流预测[J]. 计算机技术与发展, 2016, 26(11):77-81. Tan M M, Cheng X G, Zhou K, et al. Short-term traffic flow prediction based on weighted combination of ARIMA and grey model[J]. Computer Technology and Development, 2016, 26(11):77-81. (in Chinese)
[7] Castro M, Jeong Y S, Jeong M K, et al. Online-SVR for short-term traffic flow prediction under typical and atypical traffic conditions[J]. Expert Systems with Applications, 2009, 36(3):6164-6173.
[8] Liu Y, Li X, Shao X. Short-term traffic flow prediction based on lagrange support vector regression[J]. Computer & Communications, 2007:1249-1254.
[9] Lü Y, Duan Y, Kang W, et al. Traffic flow prediction with big data:a deep learning approach[J]. IEEE Transactions on Intelligent Transportation Systems, 2015, 16(2):865-873.
[10] Zhao J H, Nie Y W, Ni S J, et al. Traffic data imputation and prediction:an efficient realization of deep learning[J]. IEEE Access, 2020, 8:46713-46722.
[11] 曹博, 高茂庭. 基于LSTM的短时交通流预测研究[J]. 现代计算机, 2018(25):5-9. Cao B, Gao M T. Research on the short-term traffic flow prediction based on LSTM[J]. Modern Computer, 2018(25):5-9. (in Chinese)
[12] 戢晓峰, 戈艺澄. 基于深度学习的节假日高速公路交通流预测方法[J]. 系统仿真学报, 2020, 32(6):1164-1171. Ji X F, Ge Y C. Holiday highway traffic flow prediction method based on deep learning[J]. Journal of System Simulation, 2020, 32(6):1164-1171. (in Chinese)
[13] 刘然, 刘宇, 顾进广. 基于自适应学习率优化的AdaNet改进[J]. 计算机应用, 2020, 40(10):2804-2810. Liu R, Liu Y, Gu J G. Improved AdaNet based on adaptive learning rate optimization[J]. Journal of Computer Applications, 2020, 40(10):2804-2810. (in Chinese)
[14] Zhang D, Kabuka M. Combining weather condition data to predict traffic flow:a GRU based deep learning approach[J]. IET Intelligent Transport Systems, 2018, 12(7):578-585.
[15] Jia Y, Wu J. Rainfall-integrated traffic speed prediction using deep learning method[J]. IET Intelligent Transport Systems, 2017, 11(9):531-536.
[16] Nana H, Lei D, Li W, et al. Short-term wind speed prediction based on CNN_GRU model[C]//2019 Chinese Control and Decision Conference (CCDC). 2019:928-932.
[17] 李幼军, 黄佳进, 王海渊, 等. 基于SAE和LSTM RNN的多模态生理信号融合和情感识别研究[J]. 通信学报, 2017, 38(12):109-120. Li Y J, Huang J J, Wang H Y, et al. Study of emotion recognition based on fusion multimodal bio-signal with SAE and LSTM recurrent neural network[J]. Journal on Communications, 2017, 38(12):109-120. (in Chinese)
[18] Chen K, Chen F. Dynamic spatio-temporal graph-based CNNs for traffic flow prediction[J]. IEEE Access, 2020, 8:185136-185145.
[19] Abdeljaber O, Avci O, Kiranyaz S, et al. Real-time vibration-based structural damage detection using one-dimensional convolutional neural networks[J]. Journal of Sound and Vibration, 2017, 388:154-170.
[20] Cho H, Yoon S. Divide and conquer-based 1D CNN human activity recognition using test data sharpening[J]. Sensors, 2018, 18(4):1055-1058.
[21] Zhao Z, Zhang Y Y. A traffic flow prediction approach:LSTM with detrending[C]//2018 IEEE International Conference on Progress in Informatics and Computing (PIC). IEEE, 2018.
[22] Fu R, Zhang Z, Li L. Using LSTM and GRU neural network methods for traffic flow prediction[C]//201631st Youth Academic Annual Conference of Chinese Association of Automation (YAC). IEEE, 2016.
[23] Zhao J H, Guan X, Li X. Power allocation based on genetic simulated annealing algorithm in cognitive radio networks[J]. Chinese Journal of Electronics, 2013, 22(1):177-180.
[24] Liu L, Jiang H, He P, et al. On the variance of the adaptive learning rate and beyond[C]//International Conference on Learning Representations (ICLR). 2020.
