计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (6A): 250600172-8.doi: 10.11896/jsjkx.250600172

• 大数据&数据科学 • 上一篇    下一篇

基于多级小波分解双向Mamba的时间序列预测方法

刘鹏1, 沈吉英2, 刘东升1, 陈贵波1, 宋远威1   

  1. 1 浙江工商大学计算机科学与技术学院 杭州 310018
    2 杭州数政科技有限公司 杭州 310020
  • 出版日期:2026-06-16 发布日期:2026-06-12
  • 通讯作者: 刘东升(lds1118@zjgsu.edu.cn)
  • 作者简介:(2492047309@qq.com)
  • 基金资助:
    浙江省“尖兵”“领雁”研发计划(2026C01017,2026C01018,2025C04022,2025C01037);工业控制技术全国重点实验室开放课题(ICT2025C02)

Time Series Prediction Method Based on Multi-level Wavelet Decomposition Bidirectional Mamba

LIU Pneg1, SHEN Jiying2, LIU Dongsheng1, CHEN Guibo1, SONG Yuanwei1   

  1. 1 College of Computer Science and Technology,Zhejiang Gongshang University,Hangzhou 310018,China
    2 Hangzhou Shuzheng Technology Co.,Ltd.,Hangzhou 310020,China
  • Published:2026-06-16 Online:2026-06-12
  • About author:LIU Peng,born in 1999,postgraduate.His main research interests include time series analysis and deep learning.
    LIU Dongsheng,born in 1971,professor,master's supervisor.His main research interests include multimodal data fusion algorithms,time series prediction,anomaly detection,and other deep reinforcement learning algorithms in artificial intelligence.
  • Supported by:
    “Pioneer” and “Leading Goose” R&D Program of Zhejiang(2026C01017,2026C01018,2025C04022,2025C01037) and Open Project of State Key Laboratory of Industrial Control Technology(ICT2025C02).

PDF (PC)

48

摘要： 时间序列预测能提供对未来趋势和模式的洞察力,对于各种应用都非常重要,例如天气预报、电力负荷预测等。然而现有时间序列预测模型预测中存在模型参数大、计算复杂度高、没有充分利用数据在频域信息的问题。为了解决这一问题,提出了一种基于状态空间的新模型,用于长期时间序列预测。该模型首先采用多级小波分解将原始时序数据解耦为多个不同频带的子序列;其次,为每个子序列设计独立的双向Mamba模块以捕捉其独特的动态模式;最后,通过小波重构将各频带的预测结果精确地融合成最终预测。在ETT等7个公开数据集上的实验结果表明,该方法在多个预测长度下均取得了最优性能,相较于当前最佳的基线模型,平均MSE降低了4.12%。该方法在时间序列公共数据集上证明了其有效性和实际应用的潜力。

关键词: 时间序列预测, 多分辨率小波分解, 状态空间模型, 线性模型, 电力负荷预测

Abstract: Time series forecasting can provide insights into future trends and patterns,which is crucial for various applications.For example,weather forecast,power load forecast,etc.However,existing time series prediction models suffer from problems such as large model parameters,high computational complexity,and insufficient utilization of frequency domain information in data.To address these issues,a new model based on state space is proposed for long-term time series prediction.The model firstly uses multi-level wavelet decomposition to decouple the original time-series data into multiple sub sequences of different frequency bands.Secondly,it designs independent bidirectional Mamba modules for each subsequence to capture its unique dynamic patterns.Finally,the prediction results of each frequency band are accurately fused into the final prediction through wavelet reconstruction.Experimental results on seven publicly available datasets,including ETT,show that the method achieves optimal performance at multiple prediction lengths,with an average MSE reduction of 4.12% compared to the current best baseline model.This method has demonstrated its effectiveness and potential for practical applications on time series public datasets.

Key words: Time series prediction, Multi resolution wavelet decomposition, State space model, Linear model, Power load forecasting

中图分类号: 

  • TP391
[1] SIMS C A.Macroeconomics and reality[J].Econometrica:journal of the Econometric Society,1980:1-48.
[2] BOX G E P,JENKINS G M.Time series analysis:forecastingand control[M].San Francisco:Holden-Day,1970.
[3] SALINAS D,FLUNKERT V,GASTHAUS J,et al.DeepAR:Probabilistic forecasting with autoregressive recurrent networks[J].International Journal of Forecasting,2020,36(3):1181-1191.
[4] LIU M,ZENG A,CHEN M,et al.Scinet:Time Series Modeling and Prediction Based on Sample Convolution and Interaction [C]//Advances in Neural Information Processing Systems.2022:5816-5828.
[5] HAN K,XIAO A,WU E,et al.Transformers in Transformers[C]//Advances in Neural Information Processing Systems.2021:15908-15919.
[6] ZHOU H,ZHANG S,PENG J,et al.Information provider:Effi-cient Converter Beyond Long Range Time Series Prediction [C]//Proceedings of the AAAI Conference on Artificial Intelligence.2021:11106-11115.
[7] WU H,XU J,WANG J,et al.Autoformer:Autocorrelated Decomposition Converter for Long term Sequence Prediction [C]//Advances in Neural Information Processing Systems.2021:22419-22430.
[8] ZHOU T,MA Z,WEN Q,et al.Fedformer:Frequency enhanceddecomposition transformer for long-term sequence prediction [C]//Proceedings of the International Conference on Machine Learning.2022:27268-27286.
[9] ZHANG Y,YAN J.Crossformer:Transformer for MultivariateTime Series Prediction Using Cross Dimensional Dependence [C]//Proceedings of the 11th International Conference on Learning Representations.2023.
[10] ZENG A L,WEI C X,ZENG M X,et al.Are Transformers Effective for Time Series Forecasting?[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2023:11121-11128.
[11] YI K,ZHANG Q,FAN W,et al.Frequency-domain MLPs are More Effective Learners in Time Series Forecasting[C]//Proceedings of the 37th Conference on Neural Information Processing Systems.2023.
[12] CHEN Z,LI Y,ZHOU T,et al.HyperAttention:Long-context Attention in Near-Linear Time[J].arXiv:2310.05869,2023.
[13] LIU S,YU H,LIAO C,et al.Pyraformer:Low complexity Pyramid Attention for Remote Time Series Modeling and Prediction [C]//Proceedings of the International Conference on Learning Representations.2022.
[14] NIE Y Q,NGUYEN N H,SINTHONG P,et al.A time series is worth64 words:Long-term forecasting with transformers[C]//Proceedings of the International Conference on Learning Representations.2023.
[15] LIU Y,HU T,ZHANG H.iTransformer:Inverted transformers are effective for time series forecasting[C]//Proceedings of the International Conference on Learning Representations.2024.
[16] GU A,GOEL K,RÉ C.Efficiently Modeling Long Sequenceswith Structured State Spaces[C]//International Conference on Learning Representations.2022:1-36.
[17] GU A,DAO T.Mamba:Linear-Time Sequence Modeling with Selective State Spaces[J].arXiv:2312.00752,2023.
[18] LIEBER O,LENZ B,BATA H,et al.Jamba:a hybrid transformer-mamba language model[C]//Proceedings of the International Conference on Learning Representations.2024.
[19] WANG Z,KONG F,FENG S,et al.Is mamba effective for time series forecasting?[J].Neurocomputing,2025,619:129178.
[20] ZHAN Y,MA Z,LIU Q,et al.MambaMixer:Efficient Multi-Scale Sequence Modeling with Selective State Spaces[C]//Proceedings of the 12th International Conference on Learning Representations.2024:1-22.
[21] YANG J,WANG L,ZHANG H,et al.TimeMachine:Forecasting Time Series with Regime Switching via State Space Models[J].Neural Networks,2024:45-60.
[22] KIM T,KIM J,TAE Y,et al.Reversible instance normalization for accurate time-series forecasting against distributionshift[C]//Proceedings of the 10thInternational Conference on Learning Representations.2022.
[23] COTTER F.Uses of complex wavelets in deep convolutionalneural networks[D].Cambridge:University of Cambridge,2019.
[24] LIANG A,JIANG X,SUN Y,et al.Bi-Mamba+:Bidirectional Mamba for Time Series Forecasting[J].arXiv:2404.15772,2024.
[25] MD ATIK A,CHENG Q.Timemachine: A timeseries is worth 4 mambas for long-term forecasting[J].arXiv:2403.09898,2024.
[26] WANG Z H,KONG F H,FENG S,et al.Is mamba effectivefortime series forecasting?[J].arXiv:2403.11144,2024.
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