计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (7): 343-353.doi: 10.11896/jsjkx.250300169

• 计算机网络 • 上一篇    下一篇

基于s-TimeXer组合模型的边缘负载预测方法

石鸿凌1,2, 李锦辉1, 李成华1,2, 江小平1,2, 丁昊1,2   

  1. 1 中南民族大学电子信息工程学院 武汉 430074
    2 智能无线通信湖北省重点实验室 武汉 430074
  • 收稿日期:2025-03-31 修回日期:2025-07-27 出版日期:2026-07-15 发布日期:2026-07-10
  • 通讯作者: 江小平(jiangxp@mail.scuec.edu.cn)
  • 作者简介:(hongling.shi@scuec.edu.cn)
  • 基金资助:
    国家重点研发计划(2020YFC1522600);中南民族大学学术创新团队经费项目(XTZ24006);中央高校基本科研业务费专项资金(CZY23026)

Edge Load Prediction Method Based on s-TimeXer Combined Model

SHI Hongling1,2, LI Jinhui1, LI Chenghua1,2, JIANG Xiaoping1,2, DING Hao1,2   

  1. 1 College of Electronics and Information Engineering,South-Central Minzu University,Wuhan 430074,China
    2 Hubei Province Key Laboratory of Intelligent Wireless Communication,Wuhan 430074,China
  • Received:2025-03-31 Revised:2025-07-27 Published:2026-07-15 Online:2026-07-10
  • About author:SHI Hongling,born in 1979,postdoc-toral researcher,lecturer.His main research interests include deep learning and Internet of Thing.
    JIANG Xiaoping,born in 1974,Ph.D,associate professor.His main research interest is intelligent security.
  • Supported by:
    National Key R&D Program of China(2020YFC1522600),Academic Innovation Teams of South-Central Minzu University(XTZ24006) and Special Fund for Basic Scientific Research of Central Universities(CZY23026).

PDF (PC)

17

摘要： 边缘计算环境中的负载预测方法对于计算资源的分配管理至关重要,而边缘负载数据具有波动性、噪声性、突变性和时间依赖性等特征,因此单一预测模型难以有效提取负载数据的多维信息。针对上述问题,提出了基于s-TimeXer组合模型的边缘负载预测新方法。首先,构建FFT-SSD协同分解模块,通过快速傅里叶变换提取负载数据主周期作为奇异谱分解的窗口长度参数,增强对周期性振荡结构的捕捉能力,实现趋势项、周期项和噪声项的有效分离。然后,将负载数据作为内生变量嵌入,并将奇异谱分解的特征子序列作为外生变量嵌入,构建多维特征交互空间,通过自注意力机制捕获负载数据的时间依赖性,通过交叉注意力机制实现负载数据与奇异谱分解的特征子序列的动态交互,从而提升周期分量与趋势分量对预测目标的贡献度。同时,引入Hyperband Pruner算法实现超参数高效优化,提高预测精度。通过分解-嵌入联合优化架构,在继承TimeXer时序建模优势的同时,实现了噪声抑制与多维信息提取的协同增强。在ECW和Alibaba数据集上进行实验,结果表明,s-Time-Xer模型在预测精度上优于一系列先进的基线方法,在ECW的数据集上MSE和MAE分别降低了27.7%~63.4%和14.3%~46.5%,在Alibaba的数据集上MSE和MAE分别降低了39.7%~42.5%和18.4%~23.8%。s-TimeXer模型能够有效提升边缘负载预测的准确度,为边缘计算环境下的资源调度提供了有力支持。

关键词: 边缘计算, 负载预测, 奇异谱分解, TimeXer, Hyperband Pruner算法

Abstract: Load prediction methods in edge computing environments are crucial for the allocation and management of computing resources.Edge load data has characteristics such as volatility,noise,mutation,and time dependence.Therefore,a single prediction model is difficult to effectively extract the multi-dimensional information of load data.To address the above problems,an edge load prediction method based on thehybrid s-TimeXer model is proposed.Firstly,the FFT-SSD collaborative decomposition module is constructed.The main period of the load data is extracted by Fast Fourier Transform as the window length parameter of singular spectrum decomposition,which enhances the ability to capture periodic oscillation structure and realizes the effective separation of trend term,period term and noise term.Then,the load data is embedded as an endogenous variable,and the characteristic subsequences of singular spectrum decomposition are embedded as exogenous variables to construct a multi-dimensional feature interaction space.The time dependency of the load data is captured through the self-attention mechanism,and the dynamic interaction between the load data and the characteristic subsequences of singular spectrum decomposition is achieved through the cross-attention mechanism,thereby enhancing the contribution of the periodic component and the trend component to the prediction target.At the same time,the Hyperband Pruner algorithm is introduced to achieve efficient optimization of hyperparameters and improve prediction accuracy.Through the decomposition-embedding joint optimization architecture,while inheriting the advantages of TimeXer time series modeling,the synergistic enhancement of noise suppression and multi-dimensional information extraction is achieved.Experimental results on the ECW and Alibaba datasets demonstrate that the s-TimeXer model surpasses multiple state-of-the-art baseline methods in prediction accuracy.Specifically,on the ECW dataset,the reductions are 27.7%~63.4% for MSE and 14.3%~46.5% for MAE;on the Alibaba dataset,the reductions are 39.7%~42.5% for MSE and 18.4%~23.8% for MAE.The s-TimeXer model can effectively improve the accuracy of edge load prediction and provide strong support for resource scheduling in edge computing environments.

Key words: Edge computing, Load prediction, Singular spectrum decomposition, TimeXer, Hyperband Pruner algorithm

中图分类号: 

  • TP391
[1]SARMA A D N.Deep Learning and Edge Computing Solutions for High Performance Computing [M].Cham:Springer International Publishing,2021:47-61.
[2]LIU T,FANG L,GAO H H.Survey of Task Offloading in Edge Computing [J].Computer Science,2021,48(1):11-15.
[3]SURYA K,RAJAM V M A.Novel approaches for resourcemanagement across edge servers[J].International Journal of Networked and Distributed Computing,2023,11(1):20-30.
[4]YE X C,ZHANG H L,WANG J,et al.EMD-LSTM-based group prediction algorithm of container resource load in preprocessing molecular spectral line data[J].Journal of Jilin University(Engineering and Technology Edition),2025,55(4):1374-1383.
[5]NGUYEN C,KLEIN C,ELMROTH E.Multivariate LSTM-based location-aware workload prediction for edge data centers[C]//2019 19th IEEE/ACM International Symposium on Cluster,Cloud and Grid Computing(CCGRID).Piscataway,NJ:IEEE,2019:341-350.
[6]JIANG Q N,XU H R,CHEN Z Y,et al.Edge load Prediction with Multi-variable Spatio-Temporal Inverted Transformer[J].Journal of Chinese Computer Systems,2025,46(8):1918-1926.
[7]KUMAR N R M,ANNAPPA B,YADAV V.Efficient Kalman filter based deep learning approaches for workload prediction in cloud and edge environments[J].Computing,2024,107(1):10.
[8]WANG Y,WU H,DONG J,et al.TimeXer:Empowering Transformers for Time Series Forecasting with Exogenous Variables[J].arXiv:2402.19072,2024.
[9]CALHEIROS N R,MASOUMI E,RANJAN R,et al.Workload Prediction Using ARIMA Model and Its Impact on Cloud Applications' QoS[J].IEEE Transactions on Cloud Computing,2015,3(4):449-458.
[10]CAO L.Support vector machines experts for time series forecasting[J].Neurocomputing,2003,51:321-339.
[11]ZHONG W,ZHUANG Y,SUN J,et al.A load prediction model for cloud computing using PSO-based weighted wavelet support vector machine[J].Applied Intelligence,2018,48(11):4072-4083.
[12]TOFIGHY S,RAHMANIAN A A,GHOBAEI‐ARANI M.An ensemble CPU load prediction algorithm using a Bayesian information criterion and smooth filters in a cloud computing environment[J].Software:Practice and Experience,2018,48(12):2257-2277.
[13]DUGGAN M,MASON K,DUGGAN J,et al.Predicting host CPU utilization in cloud computing using recurrent neural networks[C]//2017 12th International Conference for Internet Technology and Secured Transactions(ICITST).Piscataway,NJ:IEEE,2017:67-72.
[14]HOCHREITER S,SCHMIDHUBER J.Long Short-Term Me-mory[J].Neural Computation,1997,9(8):1735-1780.
[15]ASHISH V,NOAM S,NIKI P,et al.Attention Is All You Need[C]//NIPS'17:Proceedings of the 31st International Confe-rence on Neural Information Processing Systems.2017:6000-6010.
[16]ZHOU H,ZHANG S,PENG J,et al.Informer:Beyond efficient transformer for long sequence time-series forecasting[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Menlo Park,CA:AAAI,2021:11106-11115.
[17]KITAEV N,KAISER U,LEVSKAYA A.Reformer:The Efficient Transformer[J].arXiv:2001.04451,2020.
[18]LIU Y,HU T,ZHANG H,et al.Itransformer:Inverted transformers are effective for time series forecasting[J].arXiv:2310.06625,2024.
[19]YOU W L,DENG L,LI R L,et al.Load Prediction Method of Cloud Resource Based on v-Informer [J].Computer Science,2024,51(12):147-156.
[20]VAUTARD R,GHIL M.Singular spectrum analysis in nonlinear dynamics,with applications to paleoclimatic time series[J].Physica D:Nonlinear Phenomena,1989,35(3):395-424.
[21]HASSANI H.Singular spectrum analysis:methodology andcomparison[J].Journal of Data Science,2007,5(2):239-257.
[22]CERNA M,HARVEY A F.The fundamentals of FFT-basedsignal analysis and measurement[R].Austin:National Instruments,2000.
[23]TRENDAFILOVA I.Singular spectrum analysis for the investigation of structural vibrations[J].Engineering Structures,2021,242:112531.
[24]HANIFI S,CAMMARONO A,ZARE-BEHTASH H.Advanced hyperparameter optimization of deep learning models for wind power prediction[J].Renewable Energy,2024,221:119700.
[25]ZHAO A,CHEN M,QUAN W,et al.A hybrid forecasting model for general hospital electricity consumption based on mixed signal decomposition[J].Energy and Buildings,2024,325:115006.
[26]LI L,JAMIESON K,DESALVO G,et al.Hyperband:A novel bandit-based approach to hyperparameter optimization[J].Journal of Machine Learning Research,2018,18(185):1-52.
[27]LI J,SELVARAJU R,GOTMARE A,et al.Align before fuse:Vision and language representation learning with momentum distillation[J].arXiv:2107.07651,2021.
[28]HUANG S,WANG Z,ZHANG H,et al.One for all:Unifiedworkload prediction for dynamic multi-tenant edge cloud platforms[C]//Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.New York:ACM,2023:788-797.
[29]CHEN L,ZHANG W,YE H.Accurate workload prediction for edge data centers:Savitzky-Golay filter,CNN and BiLSTM with attention mechanism[J].Applied Intelligence,2022,52(11):13027-13042.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发