时间二维变化建模的网络流量多步预测方法

doi:10.19665/j.issn1001-2400.20241014

摘要/Abstract

摘要：

准确预测网络流量的变化,可以帮助运营商提前进行资源分配和调度,最大程度减少网络拥塞。现有的网络流量多步预测方法难以捕获流量序列的长相关性,在多步预测任务上精度较低,基于此,提出了一种时间二维变化建模的网络流量多步预测方法。该方法首先利用门控循环单元对网络流量序列进行编码,以实现网络流量时间相关性的精准表征;然后利用网络流量周期特征对其进行重构,将一维的流量序列转化为二维,重构后的流量序列长度被压缩,特征更为集中,使得模型能够有效感知其长相关特征。最后通过新型卷积神经网络捕获重构后流量序列的二维特征,并进行加权融合得到最终的预测结果。仿真结果表明,相较于主流的网络流量多步预测方法,所提方法均方根误差至少降低约8.69%,平均绝对误差至少降低约8.96%,平均百分比误差至少降低约11.73%。实验结果说明所提方法能够有效挖掘网络流量长相关特征,在网络流量多步预测任务中具有更高的预测精度。

关键词: 预测, 网络管理, 流量预测, 时间二维变化建模

Abstract:

Accurate prediction of network traffic variations can help operators allocate resources and schedule in advance,thus minimizing network congestion.Existing multi-step prediction methods for network traffic struggle to capture the long-range dependencies in traffic sequences,resulting in a low accuracy in multi-step prediction tasks.In response,a novel method using time two-dimensional-variation modeling for multi-step network traffic prediction is proposed which first encodes the network traffic sequence using Gated Recurrent Units(GRUs) to accurately represent the temporal correlations of network traffic and then reconstructs the traffic based on its periodic characteristics,transforming the one-dimensional traffic sequence into two dimensions.The reconstructed traffic sequence has a compressed length and more concentrated features,enabling the model to effectively perceive its long-range dependencies.Finally,a novel convolutional neural network captures the two-dimensional features of the reconstructed traffic sequence and performs weighted fusion to produce the final prediction results.Simulation results show that compared to mainstream multi-step network traffic prediction methods,the proposed method reduces the root mean square error by at least 8.69%,mean absolute error by at least 8.96%,and the mean absolute percentage error by at least 11.73%.Experimental results demonstrate that the proposed method can effectively mine long-range dependencies in network traffic,achieving a higher accuracy in multi-step traffic prediction tasks.

Key words: forecasting, network management, traffic prediction, temporal 2D-variation modeling

中图分类号:

TP393.06

宋文超, 杨帆, 邢泽华, 张钰杰. 时间二维变化建模的网络流量多步预测方法[J]. 西安电子科技大学学报, 2025, 52(1): 22-36.

SONG Wenchao, YANG Fan, XING Zehua, ZHANG Yujie. Multi-step prediction method for network traffic based on temporal 2D-variation modeling[J]. Journal of Xidian University, 2025, 52(1): 22-36.

图/表 20

图1

图2

图3

图4

图5

表1

表2

表3

图6

图7

图8

图9

表4

表5

图10

图11

图12

图13

图14

表6

参考文献 35

[1]	WANG C X, YOU X H, Gao X Q, et al. On the Road to 6G:Visions,Requirements,Key Technologies and Testbeds[J]. IEEE Communications Surveys & Tutorials, 2023, 25(2):905-974.
[2]	THEODOROPOULOS T, MAROUDIS A C, VIOLOS J, et al. An Encoder-Decoder Deep Learning Approach for Multistep Service Traffic Prediction[C]// IEEE Seventh International Conference on Big Data Computing Service and Applications(BigDataService 2021).Piscataway:IEEE, 2021:33-40.
[3]	TIAN Z. Chaotic Characteristic Analysis of Network Traffic Time Series At Different Time Scales[J]. Chaos,Solitons & Fractals, 2020, 130:109412.
[4]	KO T, RAZA S M, BINH D T, et al. Network Prediction with Traffic Gradient Classification Using Convolutional Neural Networks[C]// Proceedings of the 14th International Conference on Ubiquitous Information Management and Communication(IMCOM 2020).Piscataway:IEEE, 2020:1-4.
[5]	杜秀丽, 陶帆, 范志宇, 等. 面向骨干网流量的非线性组合预测方法[J]. 小型微型计算机系统. 2023, 44(2):408-412.
	DU Xiuli, TAO Fan, FAN Zhiyu, et al. Nonlinear Combination Prediction Method for Backbone Network Traffic[J]. Journal of Miniature Microcomputer Systems, 2023, 44(2):408-412.
[6]	管业鹏, 苏光耀, 盛怡. 双向长短期记忆网络的时间序列预测方法[J]. 西安电子科技大学学报, 2024, 51(3):103-112.
	GUAN Yepeng, SU Guangyao, SHENG Yi. Time Series Prediction Method Based on the Bidirectional Long Short-Term Memory Network[J]. Journal of Xidian University, 2024, 51(3):103-112.
[7]	张可涵, 李红艳, 刘文慧, 等. 面向流量预测的时间相关图卷积网络构建方法[J]. 西安电子科技大学学报, 2023, 50(5):11-20.
	ZHANG Kehan, LI Hongyan, LIU Wenhui, et al. Construction Method of Temporal Correlation Graph Convolution Network for Traffic Prediction[J]. Journal of Xidian University, 2023, 50(5):11-20.
[8]	王跃文, 常琳, 李鸣. 基于深度学习的矿井无线网络流量长期预测研究[J]. 煤矿安全, 2021, 52(12):153-158+164.
	WANG Yuewen, CHANG Lin, LI Ming. Research on Long-Term Prediction of Mine Wireless Network Traffic Based on Deep Learning[J]. Coal Mine Safety, 2021, 52(12):153-158+164.
[9]	BI J, ZHANG X, YUAN H, et al. A Hybrid Prediction Method for Realistic Network Traffic with Temporal Convolutional Network and LSTM[J]. IEEE Transactions on Automation Science and Engineering, 2021, 19(3):1869-1879.
[10]	KONG Q, ZHANG X, ZHANG C, et al. Network Traffic Prediction:Apply the Transformer to Time Series Forecasting[J]. Mathematical Problems in Engineering, 2022, 2022(1):8424398.
[11]	WU H, HU T, LIU Y, et al. Timesnet:Temporal 2d-Variation Modeling for General Time Series Analysis[C]// The 11th International Conference on Learning Representations(ICLR 2023).Appleton:ICLR, 2023:1-23.
[12]	KE G, CHEN R S, JI S, et al. Network Traffic Prediction Based on Least Squares Support Vector Machine with Simple Estimation of Gaussian Kernel Width[J]. International Journal of Information and Computer Security, 2022, 18(1-2):1-11.
[13]	MARIET Z, KUZNETSOV V. Foundations of Sequence-to-Sequence Modeling for Time Series[C]// The 22nd International Conference on Artificial Intelligence and Statistics(AISTATS 2019). Cambridge MA: PMLR, 2019:408-417.
[14]	SHENG S, LIN K, ZHOU Y, et al. Exploring a Multi-Output Temporal Convolutional Network Driven Encoder-Decoder Framework for Ammonia Nitrogen Forecasting[J]. Journal of Environmental Management, 2023, 342:118232.
[15]	李浩阳, 贺小伟, 王宾, 等. 基于改进Informer的云计算资源负载预测[J]. 计算机工程, 2024, 50(02):43-50.
	LI Haoyang, HE Xiaowei, WANG Bin, et al. Cloud Computing Resource Load Prediction Based on Improved Informer[J]. Computer Engineering, 2024, 50(02):43-50.
[16]	葛云峰, 李红艳, 史可懿. 一种自注意力序列模型的视频流长期预测方法[J]. 西安电子科技大学学报, 2024, 51(3):88-102.
	GE Yunfeng, LI Hongyan, SHI Keyi. A Self-Attention Sequential Model for Long-Term Prediction of Video Streams[J]. Journal of Xidian University, 2024, 51(3):88-102.
[17]	LI M. Long-Range Dependence and Self-Similarity of Teletraffic with Different Protocols at the Large Time Scale of Day in the Duration of 12 Years:Autocorrelation Modeling[J]. Physica Scripta, 2020, 95(6):065222.
[18]	DU S, LI T, YANG Y, et al. An LSTM Based Encoder-Decoder Model forMultiStep Traffic Flow Prediction[C]// The 33th International Joint Conference on Neural Networks(IJCNN 2019).Piscataway:IEEE, 2019:1-8.
[19]	陈立国, 张跃冬, 耿光刚, 等. 基于GRU型循环神经网络的随机域名检测[J]. 计算机系统应用, 2018, 27(08):198-202.
	CHEN Liguo, ZHANG Yuedong, GENG Guanggang, et al. Detection of Random Domain Names Based on GRU Type Recurrent Neural Network[J]. Computer Systems Applications, 2018, 27(08):198-202.
[20]	陈宇聪, 白晓清. 考虑时序二维变化的日前市场电价预测模型[J]. 电力系统及其自动化学报, 2024, 36(07):22-29..
	CHEN Yucong, BAI Xiaoqing. Day-ahead Market Electricity Price Prediction Model Considering Temporal Two-Dimensional Variation[J]. Journal of Power Systems and Its Automation, 2024, 36(07):22-29.
[21]	LIU Z, MAO H, WU C Y, et al. A Convnet for the 2020s[C]// Proceedings ofthe 35th IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR 2022). Piscataway:IEEE, 2022:11966-11976.
[22]	SZEGEDY C, LIU W, JIA Y, et al. Going Deeper with Convolutions[C]// Proceedings ofthe 28th IEEE Conference on Computer Vision and Pattern Recognition(CVPR 2015).Piscataway:IEEE, 2015:1-9.
[23]	赵仲瑜, 唐伟忠, 张文辉, 等. 基于YOLOv5的铁路接触网异物检测模型初步研究[J]. 铁路计算机应用, 2024, 33(02):13-18.
	ZHAO Zhongyu, TANG Weizhong, ZHANG Wenhui, et al. Preliminary Study on the Model of Railway Catenary Foreign Object Detection Based on YOLOv5[J]. Railway Computer Application, 2024, 33(02):13-18.
[24]	张碧川, 刘卫东, 米浩, 等. 基于轻量化YOLOv8的安全帽检测[J]. 电脑与电信, 2024,(Z1):35-39.
	ZHANG Bichuan, LIU Weidong, MI Hao, et al. Safety Helmet Detection Based on Lightweight YOLOv8[J]. Computer & Telecommunication, 2024,(Z1):35-39.
[25]	KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Imagenet Classification with Deep Convolutional Neural Networks[C]// Proceedings of the 26th Advances in Neural Information Processing Systems(NeurIPS 2012).CA:NeurIPS, 2012:1097-1105.
[26]	SIMONYAN K, ZISSERMAN A. Very Deep Convolutional Networks for Large-Scale Image Recognition[C]// Proceedings ofthe 3rd International Conference on Learning Representations(ICLR 2015).Appleton:ICLR, 2015:1-14.
[27]	SZEGEDY C, VANHOUCKE V, IOFFE S, et al. Rethinking the Inception Architecture for Computer Vision[C]// Proceedings of the 29th IEEE Conference on Computer Vision and Pattern Recognition(CVPR 2016).Piscataway:IEEE, 2016:2818-2826.
[28]	HE K, ZHANG X, REN S, et al. Deep Residual Learning for Image Recognition[C]// Proceedings of the 29th IEEE Conference on Computer Vision and Pattern Recognition(CVPR 2016).Piscataway:IEEE, 2016:770-778.
[29]	HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely Connected Convolutional Networks[C]// Proceedings ofthe 30th IEEE Conference on Computer Vision and Pattern Recognition(CVPR 2017).Piscataway:IEEE, 2017:4700-4708.
[30]	HOWARD A G, ZHU M, CHEN B, et al. Mobilenets:Efficient Convolutional Neural Networks for Mobile Vision Applications[EB/OL].[2024-4-3]. https://arxiv.org/abs/1704.04861.
[31]	TAN M, LE Q. Efficientnet:Rethinking Model Scaling for Convolutional Neural Networks[C]// 33th International Conference on Machine Learning(ICML2016). Cambridge MA: PMLR, 2019:6105-6114.
[32]	ZAFAR N, HAQ I U, CHUGHTAI J R, et al. Applying Hybrid LSTM-GRU Model Based on Heterogeneous Data Sources for Traffic Speed Prediction in Urban Areas[J]. Sensors, 2022, 22(9):3348.
[33]	WACZYNSKA J, MARTELLI E, VALLECORSA S, et al. Convolutional LSTM Models to Estimate Network Traffic[C]// 25th International Conference on Computing in High Energy and Nuclear Physics(CHEP 2021). France: EDP Sciences, 2021, 251:02050.
[34]	ZHOU T, MA Z, WEN Q, et al. Fedformer:Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting[C]// Proceedings of the 39th International Conference on Machine Learning(ICML 2022). Cambridge MA: PMLR, 2022:27268-27286.
[35]	ZENG A, CHEN M, ZHANG L, et al. Are Transformers Effective for Time Series Forecasting?[C]// Proceedings of the 37th AAAI Conference on Artificial Intelligence(AAAI 2023). Menlo Park: AAAI, 2023:11121-11128.

	RMSE	MAE	MAPE	R²
RNN	55.627	38.452	0.224	0.535
LSTM	52.722	37.909	0.217	0.563
TCN	86.834	71.238	0.460	-0.184
Informer-Encoder	71.657	52.316	0.305	0.194
GRU	53.692	35.230	0.208	0.548

	RMSE	MAE	MAPE	R²
AlexNet	57.584	41.998	0.229	0.479
VGG-16	57.957	41.574	0.227	0.472
Inception v1	69.384	48.138	0.286	0.244
Inception v2	77.527	51.764	0.304	0.056
ResNet50	64.338	46.691	0.283	0.350
DenseNet	83.145	64.951	0.397	-0.086
MobileNet	67.748	50.305	0.305	0.279
EfficientNet	77.531	57.556	0.360	0.056
ConvNeXt	57.393	40.227	0.223	0.482

	步长	GRU	CNN- LSTM	GRU- LSTM	FEDformer	TimesNet	TCN	DLinear	GRU-Conv TimesNet
RMSE	1	21.503	21.213	20.679	21.302	21.941	21.522	20.256	22.351
	24	53.068	49.513	47.267	44.749	44.503	43.540	40.665	38.121
	48	62.968	60.388	59.784	58.185	58.450	53.022	47.829	44.007
	72	67.431	65.454	60.641	60.321	69.384	59.890	56.984	46.979
MAE	1	15.960	15.576	15.315	15.556	15.257	15.617	14.669	15.130
	24	37.695	35.462	32.729	32.106	31.731	30.439	27.637	25.364
	48	53.877	49.044	42.199	40.897	40.134	38.811	36.633	30.539
	72	61.833	56.856	50.224	51.883	48.138	44.709	46.317	31.719
MAPE	1	0.100	0.095	0.099	0.094	0.090	0.090	0.091	0.089
	24	0.305	0.268	0.287	0.224	0.176	0.162	0.201	0.143
	48	0.453	0.427	0.385	0.303	0.236	0.193	0.252	0.165
	72	0.566	0.532	0.398	0.354	0.286	0.236	0.352	0.174
R²	1	0.928	0.931	0.934	0.922	0.927	0.927	0.938	0.924
	24	0.569	0.604	0.622	0.696	0.698	0.677	0.756	0.777
	48	0.327	0.386	0.423	0.451	0.467	0.558	0.651	0.698
	72	0.064	0.133	0.187	0.208	0.244	0.375	0.499	0.653

	72	96	108	120
RMSE	46.979	47.363	50.740	53.378
MAE	31.719	32.524	36.545	37.402
MAPE	0.174	0.179	0.219	0.203
R²	0.653	0.664	0.599	0.556

	TimesNet	ConvTimesNet	GRU-ConvTimesNet
浮点数运算次数/G	33.869 0	1 714.436 7	1 718.261 8
模型参数量/M	1.276 65	59.713 9	59.717 0
训练时间/s	19.514	309.247	309.952
推理时间/ms	17.619	265.410	269.417