[1]HAO Jianlong,DU Yinhuan,XUE Lingrong,et al.A complex-valued multi-domain loss fusion network for time series forecasting[J].CAAI transactions on intelligent systems,2026,21(5):1211-1220.[doi:10.11992/tis.202511025]
Copy

A complex-valued multi-domain loss fusion network for time series forecasting

CAAI transactions on intelligent systems[ISSN 1673-4785/CN 23-1538/TP] Volume: 21 Number of periods: 2026 5 Page number: 1211-1220 Column: 学术论文—机器学习 Public date: 2026-09-05
Title:
A complex-valued multi-domain loss fusion network for time series forecasting
Author(s):
HAO Jianlong;  DU Yinhuan;  XUE Lingrong;  ZHANG Wei;  LI Ruipu
School of Information, Shanxi University of Finance and Economics, Taiyuan 030006, China
Keywords:
time series forecasting; complex-valued neural networks; loss function; complex attention mechanism; fast Fourier transform; phase aware; time-frequency fusion; dynamic gradient balancing
CLC:
TP301.6
DOI:
10.11992/tis.202511025
Abstract:
Time series forecasting possesses extensive application value in critical fields such as power systems, meteorology, and transportation. Traditional models are constrained by single temporal domain loss functions, phase information loss in the real number domain, and fixed-weight optimization imbalance, making it difficult to adapt to complex temporal characteristics. This paper proposes a complex-valued multi-domain loss fusion time series prediction network (CFuseTS). It adopts complex-valued convolution and conjugate complex-valued attention to jointly capture the amplitude and phase features of time series data, breaking the inherent limitations of real-valued modeling. Meanwhile, a multi-domain synergistic loss (MDSL) module is constructed to implement joint constraints on multivariate time series, and the dynamic gradient equalizer (DGE) dynamically adjusts the weights of each domain in real time to mitigate optimization bias from conventional fixed-weight strategies. Experiments on 8 benchmark datasets show that CFuseTS reduces mean squared error (MSE) by an average of 6.51% and a maximum of 14.78% for all predicted lengths (96/192/336/720 time steps) compared to existing advanced methods, fully validating its effectiveness and generalization ability.
References:
[1] 李毅, 彭晋卿, 廖维, 等. 一种基于时间序列的集成电力负荷预测方法研究[J]. 建筑科学, 2022, 38(10): 190-197 LI Yi, PENG Jinqing, LIAO Wei, et al. Research on an integrated model for electrical load forecasting based on time series[J]. Building science, 2022, 38(10): 190-197
[2] 骆钊, 吴谕侯, 朱家祥, 等. 基于多尺度时间序列块自编码Transformer神经网络模型的风电超短期功率预测[J]. 电网技术, 2023, 47(9): 3527-3537 LUO Zhao, WU Yuhou, ZHU Jiaxiang, et al. Wind power forecasting based on multi-scale time series block auto-encoder transformer neural network model[J]. Power system technology, 2023, 47(9): 3527-3537
[3] 梁达, 商越, 王兆芬, 等. 青海省肺结核发病与气象因素的时间序列分析[J]. 中华疾病控制杂志, 2021, 25(10): 1186-1193 LIANG Da, SHANG Yue, WANG Zhaofen, et al. Time series analysis of tuberculosis incidence and meteorological factors in Qinghai Province[J]. Chinese Journal of Disease Control and Prevention, 2021, 25(10): 1186-1193
[4] 杨文忠, 张志豪, 吾守尔·斯拉木, 等. 基于时间序列关系的GBRT交通事故预测模型[J]. 电子科技大学学报, 2020, 49(4): 615-621 YANG Wenzhong, ZHANG Zhihao, WUSHOUER Silamu, et al. GBRT traffic accident prediction model based on time series relationship[J]. Journal of University of Electronic Science and Technology of China, 2020, 49(4): 615-621
[5] 冯强, 赵建光, 杨茸, 等. 时间序列中非平稳性和波动性的建模及预测[J]. 计算机科学与探索, 2025, 19(5): 1313-1321 FENG Qiang, ZHAO Jianguang, YANG Rong, et al. Modeling and predicting time series with non-stationarity and volatility[J]. Journal of frontiers of computer science and technology, 2025, 19(5): 1313-1321
[6] 徐任超, 阎威武, 王国良, 等. 基于周期性建模的时间序列预测方法及电价预测研究[J]. 自动化学报, 2020, 46(6): 1136-1144 XU Renchao, YAN Weiwu, WANG Guoliang, et al. Time series forecasting based on seasonality modeling and its application to electricity price forecasting[J]. Acta automatica sinica, 2020, 46(6): 1136-1144
[7] 李洁, 林永峰. 基于多时间尺度RNN的时序数据预测[J]. 计算机应用与软件, 2018, 35(7): 33-37, 62 LI Jie, LIN Yongfeng. Prediction of time series data based on multi-time scale rnn[J]. Computer applications and software, 2018, 35(7): 33-37, 62
[8] 陈嘉俊, 刘波, 林伟伟, 等. 基于Transformer的时间序列预测方法综述[J]. 计算机科学, 2025, 52(6): 96-105 CHEN Jiajun, LIU Bo, LIN Weiwei, et al. Survey of transformer-based time series forecasting methods[J]. Computer science, 2025, 52(6): 96-105
[9] ZHOU Haoyi, ZHANG Shanghang, PENG Jieqi, et al. Informer: beyond efficient transformer for long sequence time-series forecasting[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver: AAAI, 2021: 11106–11115.
[10] NIE Yuqi, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with transformers[EB/OL]. (2023–03–05)[2025–11–19]. https://doi.org/10.48550/arXiv.2211.14730.
[11] LI Chengjian, SHI Zhenghao, ZHOU Liang, et al. TFFormer: a time–frequency information fusion-based CNN-transformer model for OSA detection with single-lead ECG[J]. IEEE transactions on instrumentation and measurement, 2023, 72: 1-17
[12] JIN Kaiwen, XIONG Yifeng, LOU Shuya, et al. MAFD: multiple adversarial features detector for enhanced detection of text-based adversarial examples[J]. Neural processing letters, 2024, 56(6): 251
[13] LI Zhe, LI Xuanxuan, RAO Zhongwen, et al. MTS-mixers: multivariate time series forecasting via factorized temporal and channel mixing[C]//2025 International Joint Conference on Neural Networks. Rome: IEEE, 2025: 1–8.
[14] BELTRAN A O, SUAZA-MEDINA M E, ZARAZAGA-SORIA F J, et al. Multivariate integration of time series with ML for corn price forecasting in Colombia[J]. Expert systems with applications, 2026, 299: 129822
[15] AASIM, SINGH S N, MOHAPATRA A. Repeated wavelet transform based ARIMA model for very short-term wind speed forecasting[J]. Renewable energy, 2019, 136: 758-768
[16] KUMAR DUBEY A, KUMAR A, GARC?A-D?AZ V, et al. Study and analysis of SARIMA and LSTM in forecasting time series data[J]. Sustainable energy technologies and assessments, 2021, 47: 101474
[17] SHERSTINSKY A. Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network[J]. Physica D: nonlinear phenomena, 2020, 404: 132306
[18] YU Yong, SI Xiaosheng, HU Changhua, et al. A review of recurrent neural networks: LSTM cells and network architectures[J]. Neural computation, 2019, 31(7): 1235-1270
[19] LIU Yong, HU Tengge, ZHANG Haoran, et al. iTransformer: inverted transformers are effective for time series forecasting[EB/OL]. (2024–03–14)[2025–11–19]. https://doi.org/10.48550/arXiv.2310.06625.
[20] ZHOU Tian, MA Ziqing, WANG Xue, et al. Fedformer: frequency enhanced decomposed transformer for long-term series forecasting[M]//AI for Time Series. Boca Raton: CRC Press, 2026: 10–34.
[21] YI Kun, ZHANG Qi, FAN Wei, et al. Frequency-domain MLPs are more effective learners in time series forecasting[C]//Proceedings of the 37th International Conference on Neural Information Processing Systems. New Orleans: Curran Associates Inc. , 2023: 76656–76679.
[22] WU Haixu, HU Tengge, LIU Yong, et al. TimesNet: temporal 2D-variation modeling for general time series analysis[EB/OL]. (2023–04–12)[2025–11–19]. https://doi.org/10.48550/arXiv.2210.02186.
[23] ZHANG J A, LIU Fan, MASOUROS C, et al. An overview of signal processing techniques for joint communication and radar sensing[J]. IEEE journal of selected topics in signal processing, 2021, 15(6): 1295-1315
[24] CHEN Haotian, NATSUAKI R, HIROSE A. Robust symbol detection based on quaternion neural networks in wireless polarization-shift-keying communications[J]. IEEE transactions on neural networks and learning systems, 2024, 35(11): 16064-16075
[25] LIU Mengxi, CHAI Zhuoqun, DENG Haojun, et al. A CNN-transformer network with multiscale context aggregation for fine-grained cropland change detection[J]. IEEE journal of selected topics in applied earth observations and remote sensing, 2022, 15: 4297-4306
[26] YANG Muqiao, MA M Q, LI Dongyu, et al. Complex transformer: a framework for modeling complex-valued sequence[C]//ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing. Barcelona: IEEE, 2020: 4232–4236.
[27] KIM T Y, CHO S B. Web traffic anomaly detection using C-LSTM neural networks[J]. Expert systems with applications, 2018, 106: 66-76
[28] TIAN Zhen, ZHAO W X, ZHANG Changwang, et al. EulerFormer: sequential user behavior modeling with complex vector attention[C]//Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. Washington DC: ACM, 2024: 1619–1628.
[29] KOTHAPALLY V, HANSEN J H L. Complex-valued time-frequency self-attention for speech dereverberation[EB/OL]. (2022–11–22)[2025–11–19]. https://doi.org/10.48550/arXiv.2211.12632.
[30] XIU Hao, HU Siran, CUI Yuanxin, et al. Sliding window regression method for mechanical meter reading recognition in time-series images[J]. Engineering applications of artificial intelligence, 2025, 158: 111240
[31] WANG Yulong, LIU Yushuo, DUAN Xiaoyi, et al. FilterTS: comprehensive frequency filtering for multivariate time series forecasting[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2025: 21375–21383.
[32] WANG Shiyu, WU Haixu, SHI Xiaoming, et al. TimeMixer: decomposable multiscale mixing for time series forecasting[EB/OL]. (2024–05–23)[2025–11–19]. https://doi.org/10.48550/arXiv.2405.14616.
Similar References:

Memo

-

Last Update: 2026-09-05

Copyright © CAAI Transactions on Intelligent Systems