基于稀疏高斯过程混合模型的短时交通流预测
|
摘要
交通流预测在智能交通系统中起重要作用。由于短时交通流的时变性,传统预测模型效果较差。将稀疏高斯过程混合(SGPM)模型用于短时交通流预测,并研究了隐变量后验硬划分学习算法,该算法依据最大后验估计矫正样本划分,不断迭代实现最优分组。将SGPM模型与核回归(K-R)、最小最大概率机回归(MPMR)、线性回归(L-R)以及高斯过程(GP)的预测结果对比。同时将新的学习算法与传统variational和LooCV学习算法比较。结果表明,基于新算法的SGPM模型不仅能够分模态展示预测结果、输出置信区间,且短时交通流预测均方误差可达0.047 6,训练耗时达7.121 4 s,均优于其他模型。
Forecasting of traffic flow plays an important role in intelligent transportation systems. Due to time-varying of short-term traffic flow, traditional models are less effective. A sparse Gaussian Process Mix(SGPM) model is applied to forecast short-term traffic flow. Moreover, an algorithm of posterior hard partition learning for high-efficiency hidden variables is proposed. This algorithm adopts iterative learning and using a maximizing posteriori estimation to achieve the optimal grouping of samples. The SGPM model is compared with kernel regression(K-R), minimum and maximum probability machine regression(MPMR), linear regression(L-R), and Gaussian process(GP). At the same time, the new learning algorithm is compared with traditional algorithms such as variational and LooCV learning. The results show that the SGPM model based on the proposed algorithm can display the forecasting results and confidence intervals of outputs. Its RMSE is 0.047 6, and training time is 7.121 4 s, which is superior to other models.
Forecasting of traffic flow plays an important role in intelligent transportation systems. Due to time-varying of short-term traffic flow, traditional models are less effective. A sparse Gaussian Process Mix(SGPM) model is applied to forecast short-term traffic flow. Moreover, an algorithm of posterior hard partition learning for high-efficiency hidden variables is proposed. This algorithm adopts iterative learning and using a maximizing posteriori estimation to achieve the optimal grouping of samples. The SGPM model is compared with kernel regression(K-R), minimum and maximum probability machine regression(MPMR), linear regression(L-R), and Gaussian process(GP). At the same time, the new learning algorithm is compared with traditional algorithms such as variational and LooCV learning. The results show that the SGPM model based on the proposed algorithm can display the forecasting results and confidence intervals of outputs. Its RMSE is 0.047 6, and training time is 7.121 4 s, which is superior to other models.
引文
[1] 赵晓静,舒勤,黄宏光.高速公路交通流量预测模型仿真[J].计算机仿真,2017,34(6):169-173. ZHAO Xiaojing, SHU Qin, HUANG Hongguang. Simulation of freeway traffic flow forecasting model[J]. Computer Simulation,2017,34(6):169-173. (in Chinese)
[2] 唐毅,刘卫宁,孙棣华,等.改进时间序列模型在高速公路短时交通流量预测中的应用[J].计算机应用研究,2015,32(1):146-149. TANG Yi, LIU Weining ,SUN Dihua, et al. Application of improved time series model in forecasting of short-term traffic flow for freeway[J]. Application Research of Computers,2015,32(1):146-149. (in Chinese)
[3] LIPPI M, BERTINI M, FRASCONI P. Short-term traffic flow forecasting: an experimental comparison of time-series analysis and supervised Learning[J]. IEEE Transactions on Intelligent Transportation Systems, 2013,14(2):871-882.
[4] 李晓磊,肖进丽,刘明俊.基于SARIMA模型的船舶交通流量预测研究[J].武汉理工大学学报(交通科学与工程版),2017,41(2):329-332,337. LI Xiaolei, XIAO Jinli, LIU Mingjun.Vessel traffic flow prediction based on the SARIMA model[J].Journal of Wuhan University of Technology(Transportation Science&Engineering),2017,41(2):329-332,337. (in Chinese)
[5] 梅朵,郑黎黎,冷强奎,等.基于时空GPSO-SVM的短时交通流预测[J].交通信息与安全,2017,35(2):68-74,120. MEI Duo,ZHENG Lili, LENG Qiangkui, et al. A prediction model for short-term traffic flow based on space-time GPSO-SVM[J].Journal of Transportation Information and Safety,2017,35(2):68-74,120. (in Chinese)
[6] ZHOU Yatong, ZHANG Taiyi, SUN Jiancheng .Multi-scale gaussian processes: a novel model for chaotic time series prediction[J].Chinese Physics Letters,2007,24(1):42-45.
[7] WEN Huiying, YANG Guifeng, WU Weitao. Research on the real-time prediction model of the traffic flow based on wavelet neural network[J]. Applied Mechanics & Materials, 2013,241-244(1):2088-2094.
[8] ZHANG Mingheng, Zhen Yaobao, Hui Ganglong, et al. Accurate multisteps traffic flow prediction based on SVM[J]. Mathematical Problems in Engineering, 2013,2013(6):91-109.
[9] KRANTI Kumar, MANORANJAN Parida, VINOD Kumar Katiyar. Short term traffic flow prediction in heterogeneous condition using artificial neural network[J]. Transport, 2015,30(4):397-405.
[10] 戴洪波,曾献辉.基于BP神经网络的高速公路短时交通流预测[J].智能计算机与应用,2015,5(4):36-38. DAI Hongbo, ZENG Xianhui. Highway dynamic traffic flow prediction based on BP neural network[J]. Intelligent Computer and Applications, 2015,5(4):36-38. (in Chinese)
[11] YANG Hongjun,XU Hu. Wavelet neural network with improved genetic algorithm for traffic flow time series prediction[J].OPTIK,2016,127(19):8103-8110.
[12] 李颖宏,刘乐敏,王玉全.基于组合预测模型的短时交通流预测[J].交通运输系统工程与信息,2013,13(2):34-41 LI Yinghong, LIU Lemin, WANG Yuquan.Short-term traffic flow based on combination of predictive models[J]. Journal of Transportation Systems Engineering and InformationTechnology, 2013,13(2):34-41. (in Chinese)
[13] 芮兰兰,李钦铭.基于组合模型的短时交通流量预测算法[J].电子与信息学报,2016,,38(5):1227-1233. RUI Lanlan, LI Qinming. Short-term traffic flow predicting algorithm based on combination model[J]. Journal of Electronics and Information Technology,2016,38(5):1227-1233. (in Chinese)
[14] 樊娜,赵祥模,戴明,等.短时交通流预测模型[J].交通运输工程学报,2012,12(4):114-119. FAN Na, ZHAO Xiangmo, DAI Ming, et al. Short-term traffic flow prediction model[J]. Journal of Traffic and Transportation Engineering,2012,12(4):114-119. (in Chinese)
[15] 丁恒,张卫华,郑小燕,等.基于交通预测的多态交通流信号控制[J].中国公路学报,2012,25(5):126-133. DING Heng, ZHANG Weihua, ZHENG Xiaoyan, et al. Multi-state flow signal control based on traffic prediction[J]. China Journal of Highway and Transport, 2012,25(5):126-133. (in Chinese)
[16] WANG Haiyan, WANG Youzhen. Vessel traffic flow forecasting with the combined model based on support vector machine[C]. International Conference on Transportation Information and Safety Wuhan,China: IEEE, 2015.
[17] 董春娇,邵春福,诸葛承祥.混合状态下城市快速路交通流短时预测[J].物理学报,2012,61(1):44-52. DONG Chunjiao, SHAO Chunfu, ZHUGE Chengxiang. Short-term traffic flow prediction for multi traffic states on urban expressway network[J]. Acta Physica Sinica,2012,61(1):44-52. (in Chinese)
[18] 周亚同,陈子一,马尽文.从高斯过程到高斯过程混合模型:研究与展望[J].信号处理,2016,32(8):960-972. ZHOU Yatong, CHEN Ziyi, MA Jinwen. From gaussian processes to the mixture of gaussian processes: A survey[J]. Journal of Signal Processing, 2016,32(8):960-972. (in Chinese)
[2] 唐毅,刘卫宁,孙棣华,等.改进时间序列模型在高速公路短时交通流量预测中的应用[J].计算机应用研究,2015,32(1):146-149. TANG Yi, LIU Weining ,SUN Dihua, et al. Application of improved time series model in forecasting of short-term traffic flow for freeway[J]. Application Research of Computers,2015,32(1):146-149. (in Chinese)
[3] LIPPI M, BERTINI M, FRASCONI P. Short-term traffic flow forecasting: an experimental comparison of time-series analysis and supervised Learning[J]. IEEE Transactions on Intelligent Transportation Systems, 2013,14(2):871-882.
[4] 李晓磊,肖进丽,刘明俊.基于SARIMA模型的船舶交通流量预测研究[J].武汉理工大学学报(交通科学与工程版),2017,41(2):329-332,337. LI Xiaolei, XIAO Jinli, LIU Mingjun.Vessel traffic flow prediction based on the SARIMA model[J].Journal of Wuhan University of Technology(Transportation Science&Engineering),2017,41(2):329-332,337. (in Chinese)
[5] 梅朵,郑黎黎,冷强奎,等.基于时空GPSO-SVM的短时交通流预测[J].交通信息与安全,2017,35(2):68-74,120. MEI Duo,ZHENG Lili, LENG Qiangkui, et al. A prediction model for short-term traffic flow based on space-time GPSO-SVM[J].Journal of Transportation Information and Safety,2017,35(2):68-74,120. (in Chinese)
[6] ZHOU Yatong, ZHANG Taiyi, SUN Jiancheng .Multi-scale gaussian processes: a novel model for chaotic time series prediction[J].Chinese Physics Letters,2007,24(1):42-45.
[7] WEN Huiying, YANG Guifeng, WU Weitao. Research on the real-time prediction model of the traffic flow based on wavelet neural network[J]. Applied Mechanics & Materials, 2013,241-244(1):2088-2094.
[8] ZHANG Mingheng, Zhen Yaobao, Hui Ganglong, et al. Accurate multisteps traffic flow prediction based on SVM[J]. Mathematical Problems in Engineering, 2013,2013(6):91-109.
[9] KRANTI Kumar, MANORANJAN Parida, VINOD Kumar Katiyar. Short term traffic flow prediction in heterogeneous condition using artificial neural network[J]. Transport, 2015,30(4):397-405.
[10] 戴洪波,曾献辉.基于BP神经网络的高速公路短时交通流预测[J].智能计算机与应用,2015,5(4):36-38. DAI Hongbo, ZENG Xianhui. Highway dynamic traffic flow prediction based on BP neural network[J]. Intelligent Computer and Applications, 2015,5(4):36-38. (in Chinese)
[11] YANG Hongjun,XU Hu. Wavelet neural network with improved genetic algorithm for traffic flow time series prediction[J].OPTIK,2016,127(19):8103-8110.
[12] 李颖宏,刘乐敏,王玉全.基于组合预测模型的短时交通流预测[J].交通运输系统工程与信息,2013,13(2):34-41 LI Yinghong, LIU Lemin, WANG Yuquan.Short-term traffic flow based on combination of predictive models[J]. Journal of Transportation Systems Engineering and InformationTechnology, 2013,13(2):34-41. (in Chinese)
[13] 芮兰兰,李钦铭.基于组合模型的短时交通流量预测算法[J].电子与信息学报,2016,,38(5):1227-1233. RUI Lanlan, LI Qinming. Short-term traffic flow predicting algorithm based on combination model[J]. Journal of Electronics and Information Technology,2016,38(5):1227-1233. (in Chinese)
[14] 樊娜,赵祥模,戴明,等.短时交通流预测模型[J].交通运输工程学报,2012,12(4):114-119. FAN Na, ZHAO Xiangmo, DAI Ming, et al. Short-term traffic flow prediction model[J]. Journal of Traffic and Transportation Engineering,2012,12(4):114-119. (in Chinese)
[15] 丁恒,张卫华,郑小燕,等.基于交通预测的多态交通流信号控制[J].中国公路学报,2012,25(5):126-133. DING Heng, ZHANG Weihua, ZHENG Xiaoyan, et al. Multi-state flow signal control based on traffic prediction[J]. China Journal of Highway and Transport, 2012,25(5):126-133. (in Chinese)
[16] WANG Haiyan, WANG Youzhen. Vessel traffic flow forecasting with the combined model based on support vector machine[C]. International Conference on Transportation Information and Safety Wuhan,China: IEEE, 2015.
[17] 董春娇,邵春福,诸葛承祥.混合状态下城市快速路交通流短时预测[J].物理学报,2012,61(1):44-52. DONG Chunjiao, SHAO Chunfu, ZHUGE Chengxiang. Short-term traffic flow prediction for multi traffic states on urban expressway network[J]. Acta Physica Sinica,2012,61(1):44-52. (in Chinese)
[18] 周亚同,陈子一,马尽文.从高斯过程到高斯过程混合模型:研究与展望[J].信号处理,2016,32(8):960-972. ZHOU Yatong, CHEN Ziyi, MA Jinwen. From gaussian processes to the mixture of gaussian processes: A survey[J]. Journal of Signal Processing, 2016,32(8):960-972. (in Chinese)