基于TA-Informer模型的多元长期时间序列预测研究

doi:10.3778/j.issn.1002-8331.2501-0225

摘要/Abstract

摘要： 在多元长期时间序列预测中，数据的特征冗余和长期依赖关系难以捕捉，成为影响预测精度的关键问题。为了提高多元长期时间序列预测精度，提出了一种基于TA-Informer的多元长期时间序列预测模型。模型使用时间卷积网络（TCN）对多元长时间序列进行特征提取，用于捕获长期依赖关系，将提取的特征输入自适应稀疏自注意力（ASSA）中来消除冗余特征并增强重要特征，将增强的重要特征输入Informer模块实现多元长期时间序列预测任务。实验结果表明，TA-Informer与基准模型Informer相比较，在六个公开数据集上的MSE分别下降了57.5%，25.8%，50.3%，60%，48.1%和45.2%，体现了方案的有效性和可行性。

关键词: 多元长期预测, 深度学习, 特征提取, 冗余特征, 时间卷积网络, 自适应稀疏自注意力, Informer

Abstract: In multivariate long-term time series forecasting, the feature redundancy and long-term dependency of the data are difficult to capture, which becomes a key problem affecting the forecasting accuracy. In order to improve the multivariate long-term time series prediction accuracy, a multivariate long-term time series prediction model based on TA-Informer is proposed. Firstly, the model uses temporal convolutional network (TCN) to extract features from multivariate long-term time series for capturing long-term dependencies. Then, the model feeds the extracted features into an adaptive sparse self-attention (ASSA) to eliminate redundant features and enhance the important features. Finally, the model feeds the enhanced important features into the Informer module to realize the multivariate long-term time series prediction task. The experimental results show that TA-Informer reduces the MSE on six public datasets by 57.5%, 25.8%, 50.3%, 60%, 48.1% and 45.2%, respectively, compared with the benchmark model Informer, which reflects the effectiveness and feasibility of the scheme.

Key words: multivariate long-term time series prediction, deep learning, feature extraction, redundancy feature, temporal convolutional network, adaptive sparse self-attention, Informer

王新科, 梅红岩, 赵勤, 翟心晨, 赵恩童. 基于TA-Informer模型的多元长期时间序列预测研究[J]. 计算机工程与应用, 2026, 62(8): 366-379.

WANG Xinke, MEI Hongyan, ZHAO Qin, ZHAI Xinchen, ZHAO Entong. Multivariate Long-Term Time Series Prediction Based on TA-Informer Model[J]. Computer Engineering and Applications, 2026, 62(8): 366-379.

参考文献

[1] MOHAN D P，SUBATHRA M.A comprehensive review of various machine learning techniques used in load forecasting[J].Recent Advances in Electrical & Electronic Engineering （Formerly Recent Patents on Electrical & Electronic Engineering），2023，16（3）：197-210.
[2] 娄奇鹤，刘虎，谢祥颖，等.融合级联LSTM的短期电力负荷预测[J].计算机工程与应用，2021，57（18）：275-280.
LOU Q H，LIU H，XIE X Y，et al.Short-term power load forecasting using cascaded LSTM models[J].Computer Engineering and Applications，2021，57（18）：275-280.
[3] 宋凌云，马卓源，李战怀，等.面向金融风险预测的时序图神经网络综述[J].软件学报，2024，35（8）：3897-3922.
SONG L Y，MA Z Y，LI Z H，et al.Review on temporal graph neural networks for financial risk prediction[J].Journal of Software，2024，35（8）：3897-3922.
[4] GAO M，WEI Y M.Traffic prediction based on a novel deep learning model：TPBERT[C]//Proceedings of the 6th International Conference on Artificial Intelligence and Big Data.Piscataway：IEEE，2023：241-248.
[5] 李文翰，杨琪，代治康，等.基于CEEMDAN和优化CNN-BiLSTM组合模型的短时交通流量预测[J].公路交通科技，2025，42（12）：11-23.
LI W H，YANG Q，DAI Z K,et al.Short-time traffic flow prediction based on hybrid model of CEEMDAN and optimized CNN-BiLSTM[J].Journal of Highway and TransportationResearch and Development，2025,42(12):11-23.
[6] 杨秀，闫钟宇，孙改平，等.基于多类型天气识别的光伏功率日前预测[J].现代电力，2026，43（2）：253-264.
YANG X，YAN Z Y，SUN G P,et al.Day-ahead forecasting of photovoltaic power based on multi-class weather recognition[J].Modern Electric Power，2026，43（2）：253-264.
[7] LINDEMANN B，MüLLER T，VIETZ H，et al.A survey on long short-term memory networks for time series prediction[J].Procedia CIRP，2021，99：650-655.
[8] YULE G U.On a method of investigating periodicities in disturbed series，with special reference to Wolfer??s sunspot numbers[J].Philosophical Transactions of the Royal Society of London Series A，Containing Papers of a Mathematical or Physical Character，1927，226 （1927）：267-298.
[9] BOX G E P，JENKINS G M，REINSEL G C，et al.Time series analysis：forecasting and control[M].5th ed.New York：John Wiley & Sons，2015.
[10] YANG H Y，LI X T，QIANG W H，et al.A network traffic forecasting method based on SA optimized ARIMA?BP neural network[J].Computer Networks，2021，193：108102.
[11] 谭卓昆，罗龙飞，王顺芳.基于融合特征与GA-SVM算法的脑疾病基因预测[J].计算机工程与应用，2023，59（5）：70-77.
TAN Z K，LUO L F，WANG S F.Prediction of disease gene based on fusion features and GA-SVM algorithm[J].Computer Engineering and Applications，2023，59（5）：70-77.
[12] EVERINGHAM M，VAN GOOL L，WILLIAMS C K I，et al.The pascal visual object classes（VOC） challenge[J].International Journal of Computer Vision，2010，88（2）：303-338.
[13] GU J X，WANG Z H，KUEN J，et al.Recent advances in convolutional neural networks[J].Pattern Recognition，2018，77：354-377.
[14] SALINAS D，FLUNKERT V，GASTHAUS J.DeepAR：probabilistic forecasting with autoregressive recurrent networks[J].International Journal of Forecasting，2020，36（3）：1181-1191.
[15] 张颖，李路.RF-MIP-LSTM股价预测模型[J].计算机工程与应用，2024，60（17）：272-281.
ZHANG Y，LI L.RF-MIP-LSTM stock price prediction model[J].Computer Engineering and Applications，2024，60（17）：272-281.
[16] CHO K，VAN MERRIENBOER B，GULCEHRE C，et al.Learning phrase representations using RNN encoder-decoder for statistical machine translation[EB/OL].[2024-12-28].https：//arxiv.org/abs/1406.1078.
[17] WU N，GREEN B，BEN X，et al.Deep transformer models for time series forecasting：the influenza prevalence case[EB/OL].[2024-12-29].https：//arxiv.org/abs/2001.08317.
[18] ZHOU H Y，ZHANG S H，PENG J Q，et al.Informer：beyond efficient transformer for long sequence time-series forecasting[C]//Proceedings of the 35th AAAI Conference on Artificial Intelligence.Palo Alto：AAAI，2021：11106-11115.
[19] WU H X，XU J H，WANG J M，et al.Autoformer：decomposition transformers with auto-correlation for long-term series forecasting[C]//Advances in Neural Information Processing Systems 34，2021：22419-22430.
[20] LIU S Z，YU H，LIAO C，et al.Pyraformer：low-complexity pyramidal attention for long-range time series modeling and forecasting[C]//Proceedings of the 10th International Conference on Learning Representations，2022.
[21] BAI S J，KOLTER J Z，KOLTUN V.An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[EB/OL].[2025-01-04].https：//arxiv.org/abs/1803.
01271.
[22] ZHOU S H，CHEN D S，PAN J S，et al.Adapt or perish：adaptive sparse transformer with attentive feature refinement for image restoration[C]//Proceedings of the 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2024：2952-2963.
[23] ZHOU T，MA Z Q，WEN Q S，et al.FEDformer：frequency enhanced decomposed transformer for long-term series forecasting[C]//Proceedings of the 39th International Conference on Machine Learning，2022：27268-27286.
[24] LIU Y，HU T G，ZHANG H R，et al.iTransformer：inverted transformers are effective for time series forecasting[EB/OL].[2025-01-07].https：//arxiv.org/abs/2310.06625.
[25] NIE Y Q，NGUYEN N H，SINTHONG P，et al.A time series is worth 64 words：long-term forecasting with transformers[EB/OL].[2025-01-07].https：//arxiv.org/abs/2211.14730.
[26] WANG Y X，WU H X，DONG J X，et al.TimeXer：empowering transformers for time series forecasting with exogenous variables[EB/OL].[2025-01-09].https：//arxiv.org/abs/2402.19072.
[27] LAI G K，CHANG W C，YANG Y M，et al.Modeling long- and short-term temporal patterns with deep neural networks[C]//Proceedings of the 41st International ACM SIGIR Conference on Research & Development in Information Retrieval.New York：ACM，2018：95-104.
[28] SHI J W，WANG S Q，QU P F，et al.Time series prediction model using LSTM-Transformer neural network for mine water inflow[J].Scientific Reports，2024，14：18284.