基于LSTM的城市公交车站短时客流量预测研究
|
摘要
智慧交通是智慧城市的重要组成部分,公共汽车(以下简称公交车)作为城市公共交通工具中最重要出行方式之一,不但方便了城市居民的工作和生活,而且为城市节能和环境保护提供了有效的解决方案。提高公交车站点客流量预测的准确度是智慧公交的研究内容之一。为了弥补传统时间序列模型(如ARMA和SVR)所具有的仅限单站点预测、短时间记忆等局限性,提高城市公交车站点客流量的短时预测精度,文中提出采用基于LSTM的神经网络模型对多个站点上下车客流量的长时间序列数据进行学习,从而对同一时段多个站点的客流量进行预测。试验结果表明,同时进行多站点客流量的学习能够提高预测结果的准确度,并且对抑制MSE和MAE有较好的表现,其中测试集MSE和MAE分别为3. 18人和1. 43人。基于LSTM的神经网络模型不仅能够很好发挥模型固有的长期记忆的能力,并且可以学习站点之间的潜在相关性,不仅对短时客流量预测具有明显的优势,而且拥有一定的泛化能力。使用LSTM进行多站点的公交车站客流量预测是可行的,并且较单一站点的客流量预测效果有明显提高;从客流量监测数据方面分析得出,多个公交车站点的客流量数据间存在相关性。论文成果对城市公交运营部门的快速决策和综合管理提供及时准确的数据参考具有现实意义。
Smart transport is an important part of smart city. Buses as one of the most important travel modes in urban public transport,not only facilitate the work and life of urban residents,but also provide effective solutions to urban energy conservation and environmental protection. It is one of the research contents of smart public transport to improve the forecast accuracy of passenger traffic at bus stations. In order to compensate the limitations of single station forecasting and short time memory in traditional time series models( such as ARMA and SVR) and improve the short-term prediction accuracy of urban bus station traffic volume,is proposed a neural network model based on LSTM which can learn the long time series data of traffic volume at multiple stations,so as to forecast the traffic volumes of multiple stations at the same time.The experimental result shows that( 1) the learning ability of multi-station traffic at the same time can improve the accuracy of prediction result and have better performance to suppress MSE and MAE,the MSE and MAE in the test-set are 3. 18 and 1. 43 respectively;( 2) the neural network model based on LSTM notonly can play its long-term memory ability, but also can study the potential correlation among multiple stations,it not only has obvious advantages to short-term traffic forecast but also has a certain generalization ability;( 3) using LSTM for multi-station traffic forecast is feasible,and the forecast result has improved significantly comparing with single station traffic forecast;( 4) from the traffic monitoring data analysis,the data of muti-station traffic volume have correlation. The result provides a timely and accurate data reference for rapid decision-making and integrated management of urban public transport operation departments.
Smart transport is an important part of smart city. Buses as one of the most important travel modes in urban public transport,not only facilitate the work and life of urban residents,but also provide effective solutions to urban energy conservation and environmental protection. It is one of the research contents of smart public transport to improve the forecast accuracy of passenger traffic at bus stations. In order to compensate the limitations of single station forecasting and short time memory in traditional time series models( such as ARMA and SVR) and improve the short-term prediction accuracy of urban bus station traffic volume,is proposed a neural network model based on LSTM which can learn the long time series data of traffic volume at multiple stations,so as to forecast the traffic volumes of multiple stations at the same time.The experimental result shows that( 1) the learning ability of multi-station traffic at the same time can improve the accuracy of prediction result and have better performance to suppress MSE and MAE,the MSE and MAE in the test-set are 3. 18 and 1. 43 respectively;( 2) the neural network model based on LSTM notonly can play its long-term memory ability, but also can study the potential correlation among multiple stations,it not only has obvious advantages to short-term traffic forecast but also has a certain generalization ability;( 3) using LSTM for multi-station traffic forecast is feasible,and the forecast result has improved significantly comparing with single station traffic forecast;( 4) from the traffic monitoring data analysis,the data of muti-station traffic volume have correlation. The result provides a timely and accurate data reference for rapid decision-making and integrated management of urban public transport operation departments.
引文
[1]李琦.物流需求预测中移动平均法应用分析[J].中国商贸,2011(2):139-140.LI Qi. Analysis on Application of Moving Average Method in Logistics Demand Forecasting[J]. China Business&Trade,2011(2):139-140.
[2] RICE J,ZWET E V. A Simple and Effective Method for Predicting Travel Times on Freeways[J]. IEEE Transactions on Intelligent Transportation Systems,2004,5(3):200-207.
[3] JIWON M,KIM D K,KHO S Y,et al. Travel Time Prediction Using k Nearest Neighbor Method with Combined Data from Vehicle Detector System and Automatic Toll Collection System[J]. Transportation Research Record,2011,2256:51-59.
[4] AHMED M S,COOK A R. Analysis of Freeway Traffic Time-series Data by Using Box-Jenkins Techniques[J].Transportation Research Record,1979,722:116.
[5]宋子房.公路短时车流量预测模型研究[J].科学决策,2014(4):83-94.SONG Zi-fang. Study on Short-term Traffic Volume Prediction Model[J]. Scientific Decision Making,2014(4):83-94.
[6]唐秋生,程鹏,李娜.基于GSO-BP神经网络的城市轨道交通客流量短时间预测[J].交通科技与经济,2017,19(1):1-4.TANG Qiu-sheng, CHENG Peng, LI Na. Short Time Forecasting of Passenger Flow in Urban Railway Using GSO-BPNN Method[J]. Technology&Economy in Areas of Communications,2017,19(1):1-4.
[7] RUMELHART D E,HINTON G E,WILLIAMS R J.Learning Internal Representations by Error Propagation[D]. San Diego:University of California San Diego,1985.
[8] ZHAO Z,CHEN W,WU X,et al. LSTM Network:A Deep Learning Approach for Short-term Traffic Forecast[J]. IET Intelligent Transport Systems,2017,11(2):68-75.
[9] MA X, YU H, WANG Y, et al. Large-scale Transportation Network Congestion Evolution Prediction Using Deep Learning Theory[J]. Plos One,2015,10(3):40-44.
[10] MA X, TAO Z, WANG Y, et al. Long Short-term Memory Neural Network for Traffic Speed Prediction Using Remote Microwave Sensor Data[J]. Transportation Research Part C:Emerging Technologies, 2015, 54(5):187-197.
[11] SHAO H,SOONG B H. Traffic Flow Prediction with Long Short-term Memory Networks(LSTMs)[C]//2016IEEE Region 10 Conference(TENCON). Singapore[s. n.],2016:2986-2989.
[12] WU Y K,TAN H C. Short-term Traffic Flow Forecasting with Spatial-temporal Correlation in a Hybrid Deep Learning Framework[D]. Beijing:Beijing Institute of Technology,2016.
[13]高隽.人工神经网络原理及仿真实例[M].北京:机械工业出版社,2003.GAO Jun. Principles of Artificial Neural Network and Simulation Examples[M]. Beijing:China Machine Press,2003.
[14] RUMELHART D, MCCLELLAND J. Learning Internal Representations by Error Propagation[M]. Cambridge:MIT Press,1988:318-362.
[15]朱小燕,王昱,徐伟.基于循环神经网络的语音识别模型[J].北京:计算机学报,2001,24(2):213-218.ZHU Xiao-yan, WANG Yu, XU Wei. Speech Recognition Model Based on Recurrent Neural Networks[J]. Beijing:Chinese Journal of Computers,2001,24(2):213-218.
[16] TEBELSKIS J. Speech Recognition Using Neural Networks[D]. Berlin:Siemens AG,1995.
[17]沈文炜,施惠昌.基于循环神经网络的传感器漂移补偿方法[J].计算机工程与科学,2005,27(12):95-96.SHEN Wen-wei, SHI Hui-chang. An RNN-based Approach to Compensating Sensor Drift[J]. Computer Engineering&Science,2005,27(12):95-96.
[18] BENGIO Y,DUCHARME R,VINCENT P,et al. A Neural Probabilistic Language Model[J]. Journal of Machine Learning Research, 2003, 3(2):1137-1155.
[19] 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). Wuhan:IEEE,2016:324-328.
[20] HOCHREITER S,SCHMIDHUBER J. Long Short-term Memory[J]. Neural Computation, 1997, 9(8):1735-1780.
[21] GRAVES A. Supervised Sequence Labelling with Recurrent Neural Networks[M]. Berlin:SpringerVerlag,2012.
[22]沈家军,王炜,陈峻.基于灰色马尔可夫模型的近期公交客流量预测[J].公路交通科技,2007,24(9):120-123.SHEN Jia-jun, WANG Wei, CHEN Jun. Short-term Urban Public Transit Volume Forecast Based on GrayMarkov Model[J]. Journal of Highway and Transportation Research and Development, 2007, 24(9):120-123.
[2] RICE J,ZWET E V. A Simple and Effective Method for Predicting Travel Times on Freeways[J]. IEEE Transactions on Intelligent Transportation Systems,2004,5(3):200-207.
[3] JIWON M,KIM D K,KHO S Y,et al. Travel Time Prediction Using k Nearest Neighbor Method with Combined Data from Vehicle Detector System and Automatic Toll Collection System[J]. Transportation Research Record,2011,2256:51-59.
[4] AHMED M S,COOK A R. Analysis of Freeway Traffic Time-series Data by Using Box-Jenkins Techniques[J].Transportation Research Record,1979,722:116.
[5]宋子房.公路短时车流量预测模型研究[J].科学决策,2014(4):83-94.SONG Zi-fang. Study on Short-term Traffic Volume Prediction Model[J]. Scientific Decision Making,2014(4):83-94.
[6]唐秋生,程鹏,李娜.基于GSO-BP神经网络的城市轨道交通客流量短时间预测[J].交通科技与经济,2017,19(1):1-4.TANG Qiu-sheng, CHENG Peng, LI Na. Short Time Forecasting of Passenger Flow in Urban Railway Using GSO-BPNN Method[J]. Technology&Economy in Areas of Communications,2017,19(1):1-4.
[7] RUMELHART D E,HINTON G E,WILLIAMS R J.Learning Internal Representations by Error Propagation[D]. San Diego:University of California San Diego,1985.
[8] ZHAO Z,CHEN W,WU X,et al. LSTM Network:A Deep Learning Approach for Short-term Traffic Forecast[J]. IET Intelligent Transport Systems,2017,11(2):68-75.
[9] MA X, YU H, WANG Y, et al. Large-scale Transportation Network Congestion Evolution Prediction Using Deep Learning Theory[J]. Plos One,2015,10(3):40-44.
[10] MA X, TAO Z, WANG Y, et al. Long Short-term Memory Neural Network for Traffic Speed Prediction Using Remote Microwave Sensor Data[J]. Transportation Research Part C:Emerging Technologies, 2015, 54(5):187-197.
[11] SHAO H,SOONG B H. Traffic Flow Prediction with Long Short-term Memory Networks(LSTMs)[C]//2016IEEE Region 10 Conference(TENCON). Singapore[s. n.],2016:2986-2989.
[12] WU Y K,TAN H C. Short-term Traffic Flow Forecasting with Spatial-temporal Correlation in a Hybrid Deep Learning Framework[D]. Beijing:Beijing Institute of Technology,2016.
[13]高隽.人工神经网络原理及仿真实例[M].北京:机械工业出版社,2003.GAO Jun. Principles of Artificial Neural Network and Simulation Examples[M]. Beijing:China Machine Press,2003.
[14] RUMELHART D, MCCLELLAND J. Learning Internal Representations by Error Propagation[M]. Cambridge:MIT Press,1988:318-362.
[15]朱小燕,王昱,徐伟.基于循环神经网络的语音识别模型[J].北京:计算机学报,2001,24(2):213-218.ZHU Xiao-yan, WANG Yu, XU Wei. Speech Recognition Model Based on Recurrent Neural Networks[J]. Beijing:Chinese Journal of Computers,2001,24(2):213-218.
[16] TEBELSKIS J. Speech Recognition Using Neural Networks[D]. Berlin:Siemens AG,1995.
[17]沈文炜,施惠昌.基于循环神经网络的传感器漂移补偿方法[J].计算机工程与科学,2005,27(12):95-96.SHEN Wen-wei, SHI Hui-chang. An RNN-based Approach to Compensating Sensor Drift[J]. Computer Engineering&Science,2005,27(12):95-96.
[18] BENGIO Y,DUCHARME R,VINCENT P,et al. A Neural Probabilistic Language Model[J]. Journal of Machine Learning Research, 2003, 3(2):1137-1155.
[19] 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). Wuhan:IEEE,2016:324-328.
[20] HOCHREITER S,SCHMIDHUBER J. Long Short-term Memory[J]. Neural Computation, 1997, 9(8):1735-1780.
[21] GRAVES A. Supervised Sequence Labelling with Recurrent Neural Networks[M]. Berlin:SpringerVerlag,2012.
[22]沈家军,王炜,陈峻.基于灰色马尔可夫模型的近期公交客流量预测[J].公路交通科技,2007,24(9):120-123.SHEN Jia-jun, WANG Wei, CHEN Jun. Short-term Urban Public Transit Volume Forecast Based on GrayMarkov Model[J]. Journal of Highway and Transportation Research and Development, 2007, 24(9):120-123.