融合时序卷积网络与傅里叶分解的时序预测模型

Time Series Prediction Model Combining Temporal Convolutional Network and Fourier Decomposition

摘要

针对时间序列预测任务中复杂非线性模型计算复杂度高且对噪声敏感的问题，本文提出了一种融合时序卷积网络 （Temporal Convolutional Network, TCN） 与傅里叶分解的FDTCNLinear 模型。该方法先利用傅里叶变换分离时序信号的长期趋势与季节周期项，随后采用线性层高效外推长期趋势，并通过多尺度 TCN 挖掘季节项的长短期依赖。此外，本文还提出了一种自适应动态加权混合损失（Dynamic Weighted Hybrid Loss, DWHL）函数，结合均方误差与平均绝对误差动态调整权重以降低异常值干扰。在多个公开基准数据集上的实验表明，所提模型在不同预测长度下的精度均显著优于当前 Transformer模型与线性模型。

Abstract

To tackle the high computational complexity and noise sensitivity of complex nonlinear models for time series forecasting, this paper proposes FDTCNLinear, a model integrating Temporal Convolutional Network (TCN) and the Fourier decomposition. The model first leverages the Fourier transform to decompose time-series signals into long-term trends and seasonal-cyclic components. It then adopts a linear layer for efficient extrapolation of long-term trends, and employs a multi-scale TCN to mine long- and short-term temporal dependencies hidden within seasonal components. In addition, this study designs an adaptive Dynamic Weighted Hybrid Loss (DWHL) function. By dynamically balancing the weights of Mean Squared Error (MSE) and Mean Absolute Error (MAE), this loss alleviates disturbances caused by outliers. Experiments on multiple public benchmark datasets verify that the proposed model achieves remarkably higher prediction accuracy than state-of-the-art Transformer-based models and linear baselines across all prediction horizons.

关键词

长时间序列预测 / 傅里叶分解 / 时序卷积网络 / 深度学习 / 混合模型

Key words

long-term time series prediction / Fourier decomposition / temporal convolutional network / deep learning / hybrid model

参考文献

[1] 毛远宏, 孙琛琛, 徐鲁豫, 等. 基于深度学习的时间序列预测方法综述 [J].微电子学与计算机, 2023, 40(4):8-17.

[2] BOX G E P, JENKINS G M. Some recent advances in forecasting and control[J].Journal of the Royal Statistical Society: Series C (Applied Statistics), 1968, 17(2):91-109.

[3] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J].Neural Computation, 1997, 9(8):1735-1780.

[4] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, CA, USA: Curran Associates Inc., 2017:6000-6010.

[5] 郭津延, 郑红, 杜佳宇, 等. 基于双尺度自适应令牌注意力的交通流量预测 [J].华东理工大学学报 (自然科学版), 2025, 51(6):817-826.

[6] ZHOU H, ZHANG S, PENG J, et al. Informer: Beyond efficient transformer for long sequence time-series forecasting[C]//Proceedings of the 35th AAAI Conference on Artificial Intelligence. [s.l.]: AAAI Press,2021:11106-11115.

[7] LIU S, YU H, LIAO C, et al. Pyraformer: Low-complexity pyramidal attention for long-range time series modeling and forecasting[C/OL]//International Conference on Learning Representations. [s.l.]: ICLR,2022. https://openreview.net/forum?id=0EXmFzUn5I.

[8] WU H, XU J, WANG J, et al. Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting[C]//Advances in Neural Information Processing Systems. [s.l.]: Curran Associates Inc.,2021:22419-22430.

[9] 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. Baltimore, Maryland, USA: PMLR, 2022:27268-27286.

[10] NIE Y, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: Long-term forecasting with transformers[C/OL]//Proceedings of the 11th International Conference on Learning Representations. Kigali: ICLR, 2023. https://openreview.net/forum?id=8szSgdP0fD.

[11] LIU Y, HU T, ZHANG H, et al. iTransformer: Inverted transformers are effective for time series forecasting[C/OL]//Proceedings of the 12th International Conference on Learning Representations. Vienna, Austria: ICLR, 2024. https://openreview.net/forum?id=JePfAI8fah.

[12] ZENG A, CHEN M, ZHANG L, et al. Are transformers effective for time series forecasting[C] //Proceedings of the AAAI Conference on Artificial Intelligence. Washington, DC, USA: AAAI Press, 2023:11121-11128.

[13] 文家璇, 王苗, 刘济. 基于时序分解和随机森林的时间序列多步预测算法 [J].华东理工大学学报 (自然科学版), 2023, 49(6):873-881.

[14] XIONG T, LI C, BAO Y. Seasonal forecasting of agricultural commodity price using a hybrid STL and ELM method: evidence from the vegetable market in China[J].Neurocomputing, 2018, 275:2831-2844.

[15] WEN Q, ZHANG Z, LI Y, et al. Fast Robust STL: Efficient and robust seasonal-trend decomposition for time series with complex patterns[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. [s.l.]: ACM,2020:2203-2213.

[16] GUO L, FANG W, ZHAO Q, et al. The hybrid PROPHET-SVR approach for forecasting product time series demand with seasonality[J].Computers and Industrial Engineering, 2021, 161:1075

[17] SEN R, YU F, DHILLON I. Think globally, act locally: A deep neural network approach to high-dimensional time series forecasting [C]//Proceedings of the 33rd International Conference on Neural Information Processing Systems. Vancouver, Canada: Curran Associates Inc., 2019:14084-14095.

[18] HYNDMAN R J, KOEHLER A B. Another look at measures of forecast accuracy[J].International Journal of Forecasting, 2006, 22(4):679-688.

[19] CHAI T, DRAXLER R R. Root mean square error (RMSE) or mean absolute error (MAE)?-Arguments against avoiding RMSE in the literature[J].Geoscientific Model Development, 2014, 7(3):1247-1250.

[20] BAI S, KOLTER Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling [EB/OL]. (2018-03-04). https://doi.org/10.48550/arXiv.1803.01271.

[21] California Department of Transportation. Caltrans performance measurement system (PeMS)[DB/OL]. [2025-12-18].https://pems.dot.ca.gov.

[22] Max Planck Institute for Biogeochemistry. Weather data of the max planck institute for biogeochemistry[DB/OL]. [2025-12-18].https://www.bgc-jena.mpg.de/wetter.

[23] TRINDADE A. ElectricityLoadDiagrams20112014 data set [DS/OL]. (2015-03-13) [2025-12-18].https://archive.ics.uci.edu/ml/datasets/ElectricityLoadDiagrams20112014.

[24] LAI G K. Multivariate time series data [DS/OL]. (2017-03-10) [2025-12-18].https://github.com/laiguokun/multivariate-time-series-data.

[25] WU H, HU T, LIU Y, et al. TimesNet: Temporal 2D-variation modeling for general time series analysis[C/OL]//Proceedings of the 11th International Conference on Learning Representations. Kigali: ICLR, 2023. https://openreview.net/forum?id=ju_Uqw384Oq.

[26] YUE W, LIU Y, LI H, et al. OLinear: A linear model for time series forecasting in orthogonally transformed domain[EB/OL]. (2025-05-14). https://arxiv.org/abs/2505.08550.

[27] HAN L, CHEN X Y, YE H J, et al. SOFTS: Efficient multivariate time series forecasting with series-core fusion[C]//Advances in Neural Information Processing Systems. Vancouver, Canada: Curran Associates Inc., 2024:64145-64175.

基金资助

上海市 2024 年度“科技创新行动计划”(24BC3200500, 24BC3200300)