SDformer: Fusing Series Decomposition for Superior Long-Term Time Series Forecasting

Jiayi Li; Zihang Zhang; Chao Zhang; Jun Tang; Shangce Gao

doi:10.1109/JAS.2025.125588

Volume 13 Issue 7

Jul. 2026

IEEE/CAA Journal of Automatica Sinica

JCR Impact Factor: 18.3, Top 1 (SCI Q1)

CiteScore: 28.2, Top 1% (Q1)
Google Scholar h5-index: 95， TOP 5

Turn off MathJax

Article Contents

Article Navigation > IEEE/CAA Journal of Automatica Sinica > 2026 > 13(7): 1626-1641

J. Li, Z. Zhang, C. Zhang, J. Tang, and S. Gao, “SDformer: Fusing series decomposition for superior long-term time series forecasting,” IEEE/CAA J. Autom. Sinica, vol. 13, no. 7, pp. 1626–1641, Jul. 2026. doi: 10.1109/JAS.2025.125588

Citation:

J. Li, Z. Zhang, C. Zhang, J. Tang, and S. Gao, “SDformer: Fusing series decomposition for superior long-term time series forecasting,” IEEE/CAA J. Autom. Sinica, vol. 13, no. 7, pp. 1626–1641, Jul. 2026. doi: 10.1109/JAS.2025.125588

Citation:

PDF( 17078 KB)

SDformer: Fusing Series Decomposition for Superior Long-Term Time Series Forecasting

doi: 10.1109/JAS.2025.125588

Funds: This work was partially supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI (JP25K21298, JP25K03179) and Japan Science and Technology Agency (JST) Support for Pioneering Research Initiated by the Next Generation (SPRING) (JPMJSP2145)

More Information

Author Bio:
Jiayi Li received the Ph.D. degree in engineering from the University of Toyama, Toyama, Japan, in 2026. He is currently an Assistant Professor with the College of Information Science and Engineering, Ritsumeikan University, Osaka, Japan. His current research interests include machine learning, evolutionary computation, neural networks, and bioinformatics

Zihang Zhang (Member, IEEE) received the Ph.D. degree in science and engineering from the University of Toyama, Toyama, Japan, in 2025. He is currently an Assistant Professor with the Faculty of Engineering, University of Toyama, Toyama, Japan. His current research interests include neural network for real-world applications and computational intelligence

Chao Zhang (Member, IEEE) received the Ph.D. degree at Iwate University, Japan in 2017. He is now a Specially Appointed Professor at the Faculty of Engineering, University of Toyama, Toyama, Japan. His research interests include computer vision, especially anomaly detection. He is a Member of the IEEE, IEEJ, ITE, JSAI, and IEICE

Jun Tang received the M.E. degree from Nagoya University, Japan in 1988 and continued his doctoral studies there from 1988 to 1990. His research interests include signal processing, artificial intelligence (including neural networks and machine learning), and computer operating systems, among other areas. Currently, he holds the position of Chairman and Chief Architect at Wicresoft Co., Ltd., USA. Additionally, he served as a Visiting Professor at several universities, including Beijing University and Dalian University of Technology

Shangce Gao (Senior Member, IEEE) received the Ph.D. degree in innovative life science from University of Toyama, Toyama, Japan in 2011. He is currently a Professor with the Faculty of Engineering, University of Toyama, Toyama, Japan. His current research interests include nature-inspired technologies, machine learning, and neural networks for real-world applications. He serves as an Associate Editor for many international journals, such as IEEE Transactions on Neural Networks and Learning Systems, and IEEE/CAA Journal of Automatica Sinica
Corresponding author: Shangce Gao, e-mail: gaosc@eng.u-toyama.ac.jp
Received Date: 2025-02-27
Revised Date: 2025-05-14
Accepted Date: 2025-06-05

Available Online: 2026-07-22

Abstract

Abstract

Long-term time series forecasting is crucial in numerous real-world dynamic systems and has garnered extensive research attention. In the context of time series forecasting, time series decomposition serves as an effective tool for analyzing time series data, enabling the extraction of underlying patterns and trends that enhance predictive accuracy. Despite its potential, time series decomposition has been underutilized in existing models that incorporate decomposition architectures, particularly in feature extraction. To address this gap, we propose the series decomposition encoder (SDE) block, which separates time series data into seasonal and trend components. By leveraging these decomposed elements, the SDE block enhances the model’s ability to capture essential temporal features. We substitute the initial layer of the conventional Transformer architecture with SDE, thereby introducing the series decomposition transformer (SDformer). Empirical assessments on nine benchmark datasets substantiate that our proposed SDformer attains state-of-the-art performance in long-term forecasting. The code implementation is available at the provided repository: https://github.com/Rirock/SDformer.
- Long time series forecasting,
- time series decomposition,
- time series forecasting,
- transformer

FullText(HTML)

References(51)

References

[1]	Z. Chen, M. Ma, T. Li, H. Wang, and C. Li, “Long sequence time-series forecasting with deep learning: A survey,” Inform. Fusion, vol. 97, Art. no. 101819, Sep. 2023. doi: 10.1016/j.inffus.2023.101819
[2]	J. Ji, J. Wang, C. Huang, J. Wu, B. Xu, Z. Wu, J. Zhang, and Y. Zheng, “Spatio-temporal self-supervised learning for traffic flow prediction,” in Proc. 37th AAAI Conf. Artificial Intelligence, Washington, USA, 2023, pp. 4356−4364.
[3]	C. Chen, Y. Liu, L. Chen, and C. Zhang, “Bidirectional spatial-temporal adaptive transformer for urban traffic flow forecasting,” IEEE Trans. Neural Netw. Learning Syst., vol. 34, no. 10, pp. 6913–6925, Oct. 2023. doi: 10.1109/TNNLS.2022.3183903
[4]	A. Nandi, A. De, A. Mallick, A. I. Middya, and S. Roy, “Attention based long-term air temperature forecasting network: ALTF Net,” Knowl.-Based Syst., vol. 252, Art. no. 109442, Sep. 2022. doi: 10.1016/j.knosys.2022.109442
[5]	C.-X. Zhang, J. Li, X.-F. Huang, J.-S. Zhang, and H.-C. Huang, “Forecasting stock volatility and value-at-risk based on temporal convolutional networks,” Expert Syst. Appl., vol. 207, Art. no. 117951, Nov. 2022. doi: 10.1016/j.eswa.2022.117951
[6]	X. Chen, Z. Su, L. Jin, and S. Li, “A correntropy-based echo state network with application to time series prediction,” IEEE/CAA J. Autom. Sinica, vol. 12, no. 2, pp. 425–435, Feb. 2025. doi: 10.1109/JAS.2024.124932
[7]	H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, and W. Zhang, “Informer: Beyond efficient transformer for long sequence time-series forecasting,” in Proc. 35th AAAI Conf. Artificial Intelligence, Virtual Event: 2021, pp. 11106−11115.
[8]	D. Luo and X. Wang, “ModernTCN: A modern pure convolution structure for general time series analysis,” in Proc. 12th Int. Conf. Learning Representations, Vienna, Austria, 2024.
[9]	A. Zeng, M. Chen, L. Zhang, and Q. Xu, “Are Transformers effective for time series forecasting?” in Proc. 37th AAAI Conf. Artificial Intelligence, Washington, USA, 2023, pp. 11121−11128.
[10]	A. Das, W. Kong, A. Leach, S. Mathur, R. Sen, and R. Yu, “Long-term forecasting with TiDE: Time-series dense encoder,” Trans. Mach. Learn. Res., 2023.
[11]	H. Wu, J. Xu, J. Wang, and M. Long, “Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting,” in Proc. 35th Int. Conf. Neural Information Processing Systems, Virtual Event: 2021, pp. 22419–22430.
[12]	H. Wu, T. Hu, Y. Liu, H. Zhou, J. Wang, and M. Long, “TimesNet: Temporal 2D-Variation modeling for general time series analysis,” in Proc. 11th Int. Conf. Learning Representations, Kigali, Rwanda, 2023.
[13]	Y. Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam, “A time series is worth 64 words: Long-term forecasting with transformers,” in Proc. 11th Int. Conf. Learning Representations, Kigali, Rwanda, 2023.
[14]	Y. Liu, T. Hu, H. Zhang, H. Wu, S. Wang, L. Ma, and M. Long, “iTransformer: Inverted transformers are effective for time series forecasting,” in Proc. 12th Int. Conf. Learning Representations, Vienna, Austria, 2024.
[15]	X. Cai and D. Li, “M-EDEM: A MNN-based empirical decomposition ensemble method for improved time series forecasting,” Knowl.-Based Systems, vol. 283, Art. no. 111157, Jan. 2024. doi: 10.1016/j.knosys.2023.111157
[16]	R. B. Cleveland, W. S. Cleveland, J. E. McRae, I. Terpenning, “STL: A seasonal-trend decomposition,” J. Off. Stat., vol. 6, no. 1, pp. 3–73, Mar. 1990.
[17]	H. Abbasimehr, A. Behboodi, and A. Bahrini, “A novel hybrid model to forecast seasonal and chaotic time series,” Expert Syst. Appl., vol. 239, Art. no. 122461, Apr. 2024. doi: 10.1016/j.eswa.2023.122461
[18]	K. Zhang, Q. Wen, C. Zhang, R. Cai, M. Jin, Y. Liu, et al, “Self-Supervised learning for time series analysis: Taxonomy, progress, and prospects,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 46, no. 10, pp. 6775–6794, Oct. 2024. doi: 10.1109/TPAMI.2024.3387317
[19]	T. Zhou, Z. Ma, Q. Wen, X. Wang, L. Sun, and R. Jin, “FEDformer: Frequency enhanced decomposed transformer for long-term series forecasting,” in Proc. 39th Int. Conf. Machine Learning, Baltimore, USA, 2022, pp. 27268−27286.
[20]	Z. Wang, S. Ruan, T. Huang, H. Zhou, S. Zhang, Y. Wang, L. Wang, Z. Huang, and Y. Liu, “A lightweight multi-layer perceptron for efficient multivariate time series forecasting,” Knowl.-Based Syst., vol. 288, Art. no. 111463, Mar. 2024. doi: 10.1016/j.knosys.2024.111463
[21]	R. Wang, X. Pei, J. Zhu, Z. Zhang, X. Huang, J. Zhai, and F. Zhang, “Multivariable time series forecasting using model fusion,” Inform. Sci., vol. 585, pp. 262–274, Mar. 2022. doi: 10.1016/j.ins.2021.11.025
[22]	G. P. Zhang, “Time series forecasting using a hybrid ARIMA and neural network model,” Neurocomputing, vol. 50, pp. 159–175, Jan. 2003. doi: 10.1016/S0925-2312(01)00702-0
[23]	X. Wang, Y. Kang, R. J. Hyndman, and F. Li, “Distributed ARIMA models for ultra-long time series,” Int. J. Forecast., vol. 39, no. 3, pp. 1163–1184, Jul.–Sep. 2023. doi: 10.1016/j.ijforecast.2022.05.001
[24]	H. Hewamalage, C. Bergmeir, and K. Bandara, “Recurrent neural networks for time series forecasting: Current status and future directions,” Int. J. Forecast., vol. 37, no. 1, pp. 388–427, Jan.–Mar. 2021. doi: 10.1016/j.ijforecast.2020.06.008
[25]	A. Sherstinsky, “Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network,” Phys. D: Nonlinear Phenom., vol. 404, Art. no. 132306, Mar. 2020. doi: 10.1016/j.physd.2019.132306
[26]	Y. Li, Z. Zhu, D. Kong, H. Han, and Y. Zhao, “EA-LSTM: Evolutionary attention-based LSTM for time series prediction,” Knowl.-Based Syst., vol. 181, Art. no. 104785, Oct. 2019. doi: 10.1016/j.knosys.2019.05.028
[27]	R. Dey and F. M. Salem, “Gate-variants of gated recurrent unit (GRU) neural networks,” in Proc. 60th Int. Midwest Symp. on Circuits and Systems, Boston, USA, 2017, pp. 1597−1600.
[28]	Z. Zhang, Z. Lei, M. Zhou, H. Hasegawa, and S. Gao, “Complex-valued convolutional gated recurrent neural network for ultrasound beamforming,” IEEE Trans. Neural Netw. Learn. Syst., vol. 36, no. 3, pp. 5668–5679, Mar. 2025. doi: 10.1109/TNNLS.2024.3384314
[29]	J. Chorowski, D. Bahdanau, D. Serdyuk, K. Cho, and Y. Bengio, “Attention-Based models for speech recognition,” in Proc. 29th Int. Conf. Neural Information Processing Systems, Montreal, Canada, 2015, pp. 577−585.
[30]	X. Li, Y. Liu, K. Wang, and F.-Y. Wang, “A recurrent attention and interaction model for pedestrian trajectory prediction,” IEEE/CAA J. Autom. Sinica, vol. 7, no. 5, pp. 1361–1370, Sep. 2020. doi: 10.1109/jas.2020.1003300
[31]	C. Luo, X. Zhao, Y. Sun, X. Li, and Y. Ye, “PredRANN: The spatiotemporal attention convolution recurrent neural network for precipitation nowcasting,” Knowl.-Based Systems, vol. 239, Art. no. 107900, Mar. 2022. doi: 10.1016/j.knosys.2021.107900
[32]	A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. 31st Int. Conf. Neural Information Processing Systems, Long Beach, USA, 2017, pp. 6000−6010.
[33]	A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” in Proc. 9th Int. Conf. Learning Representations, Virtual Event: 2021.
[34]	Z. Zhang, Z. Lei, M. Omura, H. Hasegawa, and S. Gao, “Dendritic learning-incorporated vision transformer for image recognition,” IEEE/CAA J. Autom. Sinica, vol. 11, no. 2, pp. 539–541, Feb. 2024. doi: 10.1109/JAS.2023.123978
[35]	S. Li, X. Jin, Y. Xuan, X. Zhou, W. Chen, Y.-X. Wang, and X. Yan, “Enhancing the locality and breaking the memory bottleneck of transformer on time series forecasting,” in Proc. 33rd Int. Conf. Neural Information Processing Systems, Vancouver, Canada, 2019, Art. no. 471.
[36]	S. Liu, H. Yu, C. Liao, J. Li, W. Lin, A. X. Liu, and S. Dustdar, “Pyraformer: Low-complexity pyramidal attention for long-range time series modeling and forecasting,” in Proc. 10th Int. Conf. Learning Representations, Virtual Event: 2022.
[37]	Y. Zhang and J. Yan, “Crossformer: Transformer utilizing cross-dimension dependency for multivariate time series forecasting,” in Proc. 11th Int. Conf. Learning Representations, Kigali, Rwanda, 2023.
[38]	P. Zeng, G. Hu, X. Zhou, S. Li, P. Liu, and S. Liu, “Muformer: A long sequence time-series forecasting model based on modified multi-head attention,” Knowl.-Based Systems, vol. 254, Art. no. 109584, Oct. 2022. doi: 10.1016/j.knosys.2022.109584
[39]	K. Fu, H. Li, and X. Shi, “CTF-former: A novel simplified multi-task learning strategy for simultaneous multivariate chaotic time series prediction,” Neural Netw., vol. 174, Art. no. 106234, Jun. 2024. doi: 10.1016/j.neunet.2024.106234
[40]	Y. Wang, H. Wu, J. Dong, G. Qin, H. Zhang, Y. Liu, Y. Qiu, J. Wang, and M. Long, “TimeXer: Empowering transformers for time series forecasting with exogenous variables,” in Proc. 37th Int. Conf. Neural Information Processing Systems, Vancouver, Canada, 2024, pp. 469−498.
[41]	S. Wang, H. Wu, X. Shi, T. Hu, H. Luo, L. Ma, J. Y. Zhang, and J. Zhou, “TimeMixer: Decomposable multiscale mixing for time series forecasting,” in Proc. 12th Int. Conf. Learning Representations, Vienna, Austria, 2024.
[42]	T. Dao and A. Gu, “Transformers are SSMs: Generalized models and efficient algorithms through structured state space duality,” in Proc. 41st Int. Conf. Machine Learning, Vienna, Austria, 2024, pp. 10041−10071.
[43]	J. Li, Z. Lei, Z. Zhang, H. Li, Y. Todo, and S. Gao, “Alternating excitation–inhibition dendritic computing for classification,” IEEE Trans. Artif. Intell., vol. 5, no. 11, pp. 5431–5441, Nov. 2024. doi: 10.1109/TAI.2024.3416236
[44]	T. Liu, J. Li, Z. Zhang, H. Yu, and S. Gao, “FD-GRNet: A dendritic-driven GRU framework for advanced stock market prediction,” IEEE Access, vol. 13, pp. 28265–28279, Jan. 2025. doi: 10.1109/ACCESS.2025.3541074
[45]	C. Lv, Y. Wang, D. Han, X. Zheng, X. Huang, and D. Li, “Efficient and effective time-series forecasting with spiking neural networks,” in Proc. 41st Int. Conf. Machine Learning, Vienna, Austria, 2024, pp. 33624−33637.
[46]	M. Jin, S. Wang, L. Ma, Z. Chu, J. Y. Zhang, X. Shi, et al., “Time-LLM: Time series forecasting by reprogramming large language models,” in Proc. 12th Int. Conf. Learning Representations, Vienna, Austria, 2024.
[47]	G. Dudek, “STD: A seasonal-trend-dispersion decomposition of time series,” IEEE Trans. Knowl. Data Eng., vol. 35, no. 10, pp. 10339–10350, Oct. 2023. doi: 10.1109/TKDE.2023.3268125
[48]	K. Bandara, C. Bergmeir, and H. Hewamalage, “LSTM-MSNet: Leveraging forecasts on sets of related time series with multiple seasonal patterns,” IEEE Trans. Neural Netw. Learn. Syst., vol. 32, no. 4, pp. 1586–1599, Apr. 2021. doi: 10.1109/TNNLS.2020.2985720
[49]	C. Deng, Y. Huang, N. Hasan, and Y. Bao, “Multi-step-ahead stock price index forecasting using long short-term memory model with multivariate empirical mode decomposition,” Inform. Sci., vol. 607, pp. 297–321, Aug. 2022. doi: 10.1016/j.ins.2022.05.088
[50]	H. He, S. Gao, T. Jin, S. Sato, and X. Zhang, “A seasonal-trend decomposition-based dendritic neuron model for financial time series prediction,” Appl. Soft Comput., vol. 108, Art. no. 107488, Sep. 2021. doi: 10.1016/j.asoc.2021.107488
[51]	B. N. Oreshkin, D. Carpov, N. Chapados, and Y. Bengio, “N-BEATS: Neural basis expansion analysis for interpretable time series forecasting,” in Proc. 8th Int. Conf. Learning Representations, Addis Ababa, Ethiopia, 2020.