基于WTC-iFCMformer模型的短时交通流预测

doi:10.3778/j.issn.1002-8331.2502-0203

摘要/Abstract

摘要： 短时交通流预测在智慧化交通的管理和调度中发挥着至关重要的作用。为更加准确地进行短时交通流预测，针对iTransformer神经网络并行结构框架在捕捉时间位置特征和隐藏时间特征方面存在的不足，引入时间季节性信息与位置编码来获取数据位置信息与时序特征；鉴于iTransformer模型在处理交通流中的短时波动和频率变化时的不足，提出了基于一维小波卷积变换网络（wavelet Transform convolution 1D，WTConv1D）的卷积提取模块，通过卷积机制增强模型对交通数据中复杂频率成分的感知和利用能力；同时考虑到短时交通流预测具有强时序性且模型计算复杂问题，引入傅里叶卷积混合机制（fast Fourier convolution mixer，FFCM），替换原始iTransformer模型的Feed-Forward模块以提取邻域时序特征、获取周期性和全局依赖特征并降低计算复杂度，从而提高WTC-iFCMformer模型预测的精度和速度。在实验部分，采用明尼苏达德卢斯大学交通实验室平台获取的交通流数据，通过对比、消融和泛化实验，在MAE、MAPE、RMSE、R2、平均收敛轮数和耗时六个指标对预测结果进行对比分析。实验结果表明，WTC-iFCMformer在多个指标上均取得了最优结果，为复杂交通场景下的实时预测与调度提供了高效且准确的解决方案。

关键词: 短时交通流预测, 深度学习, 一维小波卷积, 傅里叶卷积混合器, iTransformer, 时序建模

Abstract: Short-term traffic flow prediction plays a crucial role in the management and scheduling of intelligent transportation systems. To achieve more accurate predictions, this paper addresses the shortcomings of the iTransformer neural network??s parallel structure in capturing temporal-location features and hidden temporal features by introducing temporal seasonal information and positional encoding to acquire both spatial and temporal data characteristics. In view of the inadequacy of the iTransformer model in handling short-term fluctuations and frequency variations in traffic flow, the paper proposes a convolution extraction module based on a one-dimensional wavelet convolution transformation network (WTConv1D). This module enhances the model??s ability to perceive and utilize complex frequency components in traffic data through the WTConv1D convolution mechanism. Moreover, considering the strong temporal nature and computational complexity of short-term traffic flow prediction, the paper introduces the fast Fourier convolution mixer (FFCM) to replace the original iTransformer model??s feed-forward module. This modification allows for the extraction of neighborhood temporal features, the capture of periodicity and global dependency characteristics, and a reduction in computational complexity, thereby improving both the prediction accuracy and speed of the WTC-iFCMformer model. In the experimental section, this paper utilizes traffic flow data from the Minnesota Duluth University Traffic Laboratory platform. Through comparison, ablation, and generalization experiments, the paper analyzes and compares the prediction results in terms of six metrics: MAE, MAPE, RMSE, R2, the number of average convergence epochs, and runtime. The experimental results demonstrate that the WTC-iFCMformer model achieves optimal performance across multiple metrics, providing an efficient and accurate solution for real-time prediction and scheduling in complex traffic scenarios.

Key words: short-term traffic flow prediction, deep learning, wavelet convolution 1D, Fourier convolution mixer, iTransformer, temporal modeling

陈亮, 郭吉鑫, 雷莅平, 李巧茹. 基于WTC-iFCMformer模型的短时交通流预测[J]. 计算机工程与应用, 2026, 62(8): 355-365.

CHEN Liang, GUO Jixin, LEI Liping, LI Qiaoru. Short-Term Traffic Flow Prediction Based on WTC-iFCMformer Model[J]. Computer Engineering and Applications, 2026, 62(8): 355-365.

参考文献

[1] KARAMI Z，KASHEF R.Smart transportation planning：data，models，and algorithms[J].Transportation Engineering，2020，2：100013.
[2] LIN H C，DAI H，TSENG V S.Short-term and long-term travel time prediction using transformer-based techniques[J].Applied Sciences，2024，14（11）：4913.
[3] 熊章友，李卫军，朱晓娟，等.基于深度学习的短时交通流预测研究综述[J].计算机工程与应用，2025，61（11）：67-82.
XIONG Z Y，LI W J，ZHU X J，et al.Short-term traffic flow prediction based on deep learning[J].Computer Engineering and Applications，2025，61（11）：67-82.
[4] 韩超，宋苏，王成红.基于ARIMA模型的短时交通流实时自适应预测[J].系统仿真学报，2004，16（7）：1530-1532.
HAN C，SONG S，WANG C H.A real-time short-term traffic flow adaptive forecasting method based on ARIMA model[J].Journal of System Simulation，2004，16（7）：1530-1532.
[5] 李敏.基于GJR-GARCH模型和一种集成学习方法的短时交通流预测研究[D].太原：太原理工大学，2019.
LI M.Investigation of short?term traffic flow prediction based on GJR?GARCH model with an ensemble learning method[D].Taiyuan：Taiyuan University of Technology，2019.
[6] 王均，关伟.基于Kalman滤波的城市环路交通流短时预测研究[J].交通与计算机，2006，24（5）：16-19.
WANG J，GUAN W.Short-term traffic forecasting of city ring road based on Kalman filter[J].Computer and Communications，2006，24（5）：16-19.
[7] 程山英.基于K近邻非参数回归的短时交通流预测模型研究[J].数字技术与应用，2017，35（4）：130-132.
CHENG S Y.Research on short-term traffic flow forecasting model based on K-nearest neighbor nonparametric regression[J].Digital Technology and Application，2017，35（4）：130-132.
[8] 杨兆升，王媛，管青.基于支持向量机方法的短时交通流量预测方法[J].吉林大学学报（工学版），2006，36（6）：881-884.
YANG Z S，WANG Y，GUAN Q.Short-term traffic flow prediction method based on SVM[J].Journal of Jilin University （Engineering and Technology Edition），2006，36（6）：881-884.
[9] TING T J，LI X C，SANNER S，et al.Revisiting random forests in a comparative evaluation of graph convolutional neural network variants for traffic prediction[C]//Proceedings of the 2021 IEEE International Intelligent Transportation Systems Conference.Piscataway：IEEE，2021：1259-1265.
[10] 王建，邓卫，赵金宝.基于贝叶斯网络多方法组合的短时交通流量预测[J].交通运输系统工程与信息，2011，11（4）：147-153.
WANG J，DENG W，ZHAO J B.Short-term freeway traffic flow prediction based on multiple methods with Bayesian network[J].Journal of Transportation Systems Engineering and Information Technology，2011，11（4）：147-153.
[11] CAO X F，ZHONG Y H，ZHOU Y，et al.Interactive temporal recurrent convolution network for traffic prediction in data centers[J].IEEE Access，2018，6：5276-5289.
[12] DU B W，PENG H，WANG S Z，et al.Deep irregular convolutional residual LSTM for urban traffic passenger flows prediction[J].IEEE Transactions on Intelligent Transportation Systems，2020，21（3）：972-985.
[13] 徐钽，谢开贵，王宇，等.基于TCN-Wpsformer混合模型的超短期风电功率预测[J].电力自动化设备，2024，44（8）：54-61.
XU T，XIE K G，WANG Y，et al.Ultra-short-term wind power forecasting based on TCN?Wpsformer hybrid model[J].Electric Power Automation Equipment，2024，44（8）：54-61.
[14] YANG Y J，ZHANG J L，YANG L X，et al.Short-term passenger flow prediction for multi-traffic modes：a Transformer and residual network based multi-task learning method[J].Information Sciences，2023，642：119144.
[15] 杨国亮，习浩，龚家仁，等.基于Transformer的短时交通流时空预测[J].计算机应用与软件，2024，41（3）：169-173.
YANG G L，XI H，GONG J R，et al.Short term traffic flow forecasting based on trasnsformer[J].Computer Applications and Software，2024，41（3）：169-173.
[16] 廖梦媛，戎荣，张梦琦，等.基于GCN-Transformer模型的短时交通量确定性预测[J].交通技术，2024（4）：234-243.
LIAO M Y，RONG R，ZHANG M Q，et al.Deterministic prediction of short-term traffic volume based on GCN-transformer model[J].Open Journal of Transportation Technologies，2024（4）：234-243.
[17] SHAO Z Q，BELL M G H，WANG Z，et al.CCDSReFormer：traffic flow prediction with a criss-crossed dual-stream enhanced rectified transformer model[EB/OL].[2025-02-04].https：//arxiv.org/abs/2403.17753.
[18] ZHANG J H，JIN J C，TANG J J，et al.FPTN：fast pure transformer network for traffic flow forecasting[C]//Proceedings of the 32nd International Conference on Artificial Neural Networks and Machine Learning.Cham：Springer，2023：382-393.
[19] WEI S W，YANG Y，LIU D H，et al.Transformer-based spatiotemporal graph diffusion convolution network for traffic flow forecasting[J].Electronics，2024，13（16）：3151.
[20] 谷小兵，牛少彰，王茂森，等.基于Transformer的时间序列插补技术研究[J].图像与信号处理，2024，13（2）：151-162.
GU X B，NIU S Z，WANG M S，et al.Research on transformer-based time series imputation technique[J].Journal of Image and Signal Processing，2024，13（2）：151-162.
[21] JIA W H，GUAN S P，XUE Y W.TL-iTransformer：revolutionizing sea surface temperature prediction through iTransformer and transfer learning[J].Earth Science Informatics，2024，17（5）：4847-4857.
[22] LIU Y，HU T G，ZHANG H R，et al.iTransformer：inverted transformers are effective for time series forecasting[C]//Proceedings of the 12th International Conference on Learning Representations，2024.
[23] 秦严严.交通流分析理论[M].北京：人民交通出版社，2023.
QIN Y Y.Theory of traffic flow analysis[M].Beijing：China Communications Press，2023.
[24] FINDER S E，AMOYAL R，TREISTER E，et al.Wavelet convolutions for large receptive fields[C]//Proceedings of the 18th European Conference on Computer Vision，2024：363-380.
[24] ZHAO L，SONG Y J，ZHANG C，et al.T-GCN：a temporal graph convolutional network for traffic prediction[J].IEEE Transactions on Intelligent Transportation Systems，2020，21（9）：3848-3858.
[25] POPEL M，BOJAR O.Training tips for the Transformer model[J].The Prague Bulletin of Mathematical Linguistics，2018，110（1）：43-70.