基于SARIMA-GPR模型的短时交通流预测应用研究
Research on the Application of Short-Term Traffic Flow Prediction Based on SARIMA-GPR Model
DOI: 10.12677/ORF.2022.122040, PDF,    国家自然科学基金支持
作者: 王飞云：贵州大学数学与统计学院，贵州 贵阳；胡 尧：贵州大学数学与统计学院，贵州 贵阳；贵州大学公共大数据国家重点实验室，贵州 贵阳
关键词: 交通流；SARIMA模型；GPR模型；组合模型；预测；Traffic Flow； SARIMA Model； GPR Model； Combined Model； Prediction
摘要: 交通流量数据具有周期性、不平稳性、复杂性等特点，若使用单一模型对其进行预测，则预测效果不是很好，因此提出一种组合的SARIMA-GPR模型。SARIMA (Seasonal Autoregressive Integrated Moving Average)模型与GPR (Gaussian Process Regression)模型分别很好拟合交通流量的线性部分与非线性部分，且GPR模型考虑到数据的噪声，能更好地抓取到数据信息。对原数据进行特征提取与分析，训练SARIMA模型与GPR模型，得到两个预测模型，根据模型的MAE得到两个模型的权重值，得到最终的预测值。将该组合模型与SARIMA、GPR、SVM、SARIMA-SVM组合模型进行预测效果对比，实验结果表明，SARIMA-GPR模型预测效果要优于单一模型，预测结果平均绝对百分比误差(MAPE)减少到4.51%，预测结果更接近真实数据。
Abstract: Traffic flow data have the characteristics of periodicity, instability and complexity. If a single model is used to predict it, the prediction effect is not very good. Therefore, a combined SARIMA-GPR model is proposed. The SARIMA (Seasonal Autoregressive Integrated Moving Average) model and the GPR (Gaussian Process Regression) model fit the linear part and the nonlinear part of the traffic flow well respectively, and the GPR model takes into account the noise of the data and can better capture the data information. Perform feature extraction and analysis on the original data, train the SARIMA model and the GPR model, and obtain two prediction models. According to the MAE of the model, the weight values of the two models are obtained, and the final prediction value is obtained. The combined model is compared with SARIMA, GPR, SVM, and SARIMA-SVM combined model. The experimental results show that the prediction effect of SARIMA-GPR model is better than that of a single model, and the mean absolute percentage error (MAPE) of the prediction results is reduced to 4.51%, the prediction effect is closer to the real data.
文章引用：王飞云, 胡尧. 基于SARIMA-GPR模型的短时交通流预测应用研究[J]. 运筹与模糊学, 2022, 12(2): 388-396. https://doi.org/10.12677/ORF.2022.122040

参考文献

[1] 2019年贵阳市国民经济和社会发展统计公报[Z]. 贵阳市统计局, 2020. http://tjj.guiyang.gov.cn/2020_zwgk
[2] Polson, N.G. and Sokolov, V.O. (2017) Deep Learning for Short-Term Traffic Flow Prediction. Transportation Research Part C: Emerging Technologies, 79, 1-17. [Google Scholar] [CrossRef] 
[3] Ryu, U., Wang, J., Kim, T., et al. (2018) Construction of Traffic State Vector Using Mutual Information for Short-Term Traffic Flow Prediction. Transportation Research Part C: Emerging Technologies, 96, 55-71. [Google Scholar] [CrossRef] 
[4] Emami, A., Sarvi, M. and Bagloee, S.A. (2020) Short-Term Traffic Flow Prediction Based on Faded Memory Kalman Filter Fusing Data from Connected Vehicles and Bluetooth Sensors. Simulation Modelling Practice and Theory, 102, Article ID: 102025. [Google Scholar] [CrossRef] 
[5] 胡伍生, 吕楚男, 夏晓明. 基于神经网络的短期交通流预测模型[J]. 现代测绘, 2020, 43(5): 10-13
[6] 温美玲, 路鹏远, 蔡林, 等. 基于轨迹大数据的交通拥堵评估和预测[J]. 数字制造科学, 2021, 19(1): 77-80.
[7] 孙越, 宋晓宇, 金莉婷, 刘童.基于ARMA-LSTM组合模型的铁路客流量预测[J]. 计算机应用与软件, 2021, 38(12): 262-267+273.
[8] 温惠英, 曹正. 基于时延特性的短时动态交通流预测模型研究[J]. 计算机仿真, 2021, 38(6): 93-97.
[9] 张国赟, 金辉. 基于改进ARIMA模型的城市轨道交通短时客流预测研究[J]. 计算机应用与软件, 2022, 39(1): 339-344.
[10] Yaya, O.S. and Fashae, O.A. (2015) Seasonal Fractional Integrated Time Series Models for Rainfall Data in Nigeria. Theoretical and Applied Climatology, 120, 99-108. [Google Scholar] [CrossRef] 
[11] Thomas, J., Brian, J. and William, I. (2018) The Design and Analysis of Computer Experiments. Springer Science + Business Media, LLC, Berlin, 115-118.
[12] Sugasawa, S. (2020) Robust Empirical Bayes Small Area Estimation with Density Power Divergence. Biometrika, 107, 467-480. [Google Scholar] [CrossRef] 
[13] Liu, Z., Jiang, P., Zhang, L., et al. (2020) A Combined Forecasting Model for Time Series: Application to Short-Term Wind Speed Forecasting. Applied Energy, 259, Article ID: 114137. [Google Scholar] [CrossRef] 
[14] 李键, 牛峰, 黄晓艳, 等. 永磁同步电机有限控集模型预测电流控制预测误差分析[J]. 电机与控制学报, 2019, 23(4):1-7.
[15] Dridi, N. and Hadzagic, M. (2018) Akaike and Bayesian Information Criteria for Hidden Markov Models. IEEE Signal Processing Letters, 26, 302-306. [Google Scholar] [CrossRef] 

为你推荐