吉林大学学报(工学版) ›› 2024, Vol. 54 ›› Issue (9): 2531-2539.doi: 10.13229/j.cnki.jdxbgxb.20221455

• 交通运输工程·土木工程 • 上一篇    

CapsNet融合D-BiLSTM的区域复杂路网交通速度预测

曹洁1,2(),苏广1,张红1(),李鹏辉1   

  1. 1.兰州理工大学 计算机与通信学院，兰州 730050
    2.兰州城市学院 信息工程学院，兰州 730070
  • 收稿日期:2022-11-14 出版日期:2024-09-01 发布日期:2024-10-28
  • 通讯作者: 张红 E-mail:caoj@lut.edu.cn;zhanghong@lut.edu.cn
  • 作者简介:曹洁（1966-），女，教授，硕士.研究方向：模式识别理论及应用，智能交通系统.E-mail：caoj@lut.edu.cn
  • 基金资助:
    甘肃省重点研发计划项目(23YFGA0063);国家自然科学研究基金项目(62363022)

Traffic speed prediction of regional complex road networks integrating CapsNet with D-BiLSTM

Jie CAO1,2(),Guang SU1,Hong ZHANG1(),Peng-hui LI1   

  1. 1.School of Computer and Communication，Lanzhou University of Technoogy，Lanzhou 730050，China
    2.School of Information Engineening，Lanzhou City University，Lanzhou 730070，China
  • Received:2022-11-14 Online:2024-09-01 Published:2024-10-28
  • Contact: Hong ZHANG E-mail:caoj@lut.edu.cn;zhanghong@lut.edu.cn

RICH HTML

 1   

PDF (PC)

38

摘要：

针对交通模式复杂和动态的时空相关性导致现有预测方法在结构深度和预测尺度方面不足以学习交通演变的问题，提出了一种结合胶囊网络（CapsNet）和深层双向LSTM（D-BiLSTM）的深度学习模型。该模型采用CapsNet识别路网的空间拓扑结构并提取空间特征，融合D-BiLSTM网络，同时考虑交通状态的前向和后向依赖关系，捕获不同历史时期的双向时间相关性，对目标区域内大规模复杂路网的交通进行预测。在真实交通路网速度数据集上进行的实验表明：提出模型的预测精度平均提高了10%以上，优于其他方法，在区域复杂路网的交通预测中具有较高的预测精度和良好的鲁棒性。

关键词: 胶囊网络, 深层双向LSTM, 复杂路网, 后向依赖, 交通速度预测

Abstract:

Due to the complex and dynamic spatio-temporal correlation of traffic patterns leads to the inadequacy of existing methods to learn traffic evolution in terms of structural depth and prediction scale. a Deep learning model combining CapsNet and deep bi-directional LSTM （D-BiLSTM） was proposed. This model was used to identify the spatial topology of road networks and extract spatial features using CapsNet,was fused with the D-BiLSTM network, taking into account both the forward and backward dependencies of traffic states, and capturing the bi-directional temporal correlations of different historical periods, to forecast traffic on large-scale complex road networks in the target region. Experiments conducted on real traffic road network speed datasets show that the prediction accuracy of the proposed model is improved by more than 10% on average， outperforming other methods， with high prediction accuracy and good robustness in traffic prediction of regional complex road networks.

Key words: capsule network, deep bidirectional LSTM, complex road network, backward dependence, traffic speed forecast

中图分类号: 

  • TP183

图1

交通路网数据到图像的转换"

图2

CapsNet模型结构图"

图3

注意力模块结构图"

图4

LSTM体系结构图"

图5

BiLSTM体系结构图"

图6

CapsNet+D-BiLSTM模型结构图"

表1

CapsNet+D-BiLSTM模型结构参数"

层名参数输出大小
输入-164×148×10
卷积1

卷积核=3×3

通道=128

步长=2

87×73×1281 280
卷积2

卷积核=3×3

通道=128

步长=2

40×36×128147 584
注意力-
主胶囊层

卷积核=3×3

通道=128

步长=4

9×8×128147 584
胶囊维度=81 152×8
高级胶囊层

高级胶囊=30

胶囊维度=16

1 152×8

9 216

0
双向LSTM隐藏单元=200（None，200）7 453 600
舍弃层0.2（None，200）0
全连接层-（None，278）55 875
参数总数7 808 230

表2

CapsNet+D-BiLSTM模型与其他模型性能对比"

模型2 min10 min20 min
MAEMAPE/%MAEMAPE/%MAEMAPE/%
CapsNet4.517 222.174 94.849 523.430 45.792 026.242 6
BiLSTM5.703 230.863 15.795 230.885 66.318 834.375 4
CNN+LSTM4.783 024.692 45.057 125.002 05.473 326.003 8
CapsNet+LSTM4.313 521.636 34.764 425.038 14.906 425.806 5
CNN+BiLSTM4.597 521.565 44.971 624.721 05.015 125.051 5
CapsNet+D-BiLSTM4.154 620.663 54.748 722.616 54.890 423.465 9

图7

各模型不同时间步长下的MAE结果"

图8

各模型不同时间步长下的MAPE结果"

图9

各模型训练时间对比图"

图10

不同时滞阶段下的MAE箱线图"

1 James J Q. Citywide traffic speed prediction: a geometric deep learning approach[J]. Knowledge-Based Systems, 2021, 212: No.106592.
2 Wu P, Huang Z, Pian Y, et al. A combined deep learning method with attention-based LSTM model for short-term traffic speed forecasting[J]. Journal of Advanced Transportation, 2020, 2020: No.8863724.
3 Navarro-Espinoza A, López-Bonilla O R, García-Guerrero E E, et al. Traffic flow prediction for smart traffic lights using machine learning algorithms[J]. Technologies, 2022, 10(1): 5.
4 Huang Z, Xu L, Lin Y. Multi-stage pedestrian positioning using filtered WiFi scanner data in an urban road environment[J]. Sensors, 2020, 20(11):No. 20113259.
5 Yang H, Li X, Qiang W, et al. A network traffic forecasting method based on SA optimized ARIMA–BP neural network[J]. Computer Networks, 2021, 193: No. 108102.
6 Kumar P B, Hariharan K. Time series traffic flow prediction with hyper-parameter optimized ARIMA models for intelligent transportation system[J]. Journal of Scientific and Industrial Research (JSIR), 2022, 81(4): 408-415.
7 Ghosh B, Dauwels J. Comparison of different Bayesian methods for estimating error bars with incident duration prediction[J]. Journal of Intelligent Transportation Systems, 2022, 26(4): 420-431.
8 Xu D, Wang Y, Peng P, et al. Real-time road traffic state prediction based on kernel-KNN[J]. Transportmetrica A: Transport Science, 2020, 16(1): 104-118.
9 Zhang L, Ma J, Liu X, et al. A novel support vector machine model of traffic state identification of urban expressway integrating parallel genetic and C-means clustering algorithm[J]. Tehnički vjesnik, 2022, 29(3): 731-741.
10 Toan T D, Truong V H. Support vector machine for short-term traffic flow prediction and improvement of its model training using nearest neighbor approach[J]. Transportation Research Record, 2021, 2675(4): 362-373.
11 Feng B, Xu J, Zhang Y, et al. Multi-step traffic speed prediction based on ensemble learning on an urban road network[J]. Applied Sciences, 2021, 11(10):No. 11104423.
12 Bao X, Jiang D, Yang X, et al. An improved deep belief network for traffic prediction considering weather factors[J]. Alexandria Engineering Journal, 2021, 60(1): 413-420.
13 Nguyen T, Nguyen G, Nguyen B M. EO-CNN: an enhanced CNN model trained by equilibrium optimization for traffic transportation prediction[J]. Procedia Computer Science, 2020, 176: 800-809.
14 Sha S, Li J, Zhang K, et al. RNN-based subway passenger flow rolling prediction[J]. IEEE Access, 2020, 8: 15232-15240.
15 Ma X, Zhong H, Li Y, et al. Forecasting transportation network speed using deep capsule networks with nested LSTM models[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 22(8): 4813-4824.
16 Abduljabbar R L, Dia H, Tsai P W. Unidirectional and bidirectional LSTM models for short-term traffic prediction[J]. Journal of Advanced Transportation, 2021, 11(1): No.5589075.
17 Wang S, Zhao J, Shao C, et al. Truck traffic flow prediction based on LSTM and GRU methods with sampled GPS data[J]. IEEE Access, 2020, 8: 208158-208169.
18 Van L T, Nguyen Q H, Le T D T. Emotion recognition with capsule neural network[J]. Computer Systems Science and Engineering, 2022, 41(3): 1083-1098.
19 Sabour S, Frosst N, Hinton G E. Dynamic routing between capsules[J]. Advances in Neural Information Processing Systems, 2017, 30:3856-3866.
[1] 温晓岳,钱国敏,孔桦桦,缪月洁,王殿海. TrafficPro：一种针对城市信控路网的路段速度预测框架[J]. 吉林大学学报(工学版), 2024, 54(8): 2214-2222.
[2] 王生生,李晨旭,王翔宇,姚志林,刘一申,吴佳倩,杨晴然. 基于改进残差胶囊网络和麻雀搜索的脑瘤图像分类[J]. 吉林大学学报(工学版), 2022, 52(11): 2653-2661.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 《吉林大学学报(工学版)》编辑部
地址：长春市人民大街5988号　电话：+86-431-85095297　传真：+86-431-85094074　E-mail: xbgxb@jlu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发