面向长期时间序列预测的多局部增强线性解码器研究

doi:10.3778/j.issn.1002-8331.2406-0366

摘要/Abstract

摘要： 在长期时间序列预测（long-term time series forecasting，LTSF）领域，采用编码器-解码器架构的深度学习模型展现出了卓越的性能。目前，虽然编码器从输入的历史序列中能够提取深层次的时间变化特征，捕获时间序列内部的周期性、趋势性以及局部信息相关性，但是解码器多依赖于编码器输出的全局特征，对编码器提取的局部特征利用不充分，限制了模型的预测性能。为充分挖掘和利用局部特征，提出了一种多局部增强线性解码器（multiple local augmented linear decoders，MLAD），通过引入局部特征增强机制（local feature augmented mechanism，LFAM），在编码器生成的特征序列上进行滑动，并将提取的局部特征与原始历史序列融合，从而生成多个局部增强序列，然后通过计算所有的局部增强序列对应预测序列的平均值来确定预测结果。在7个公开数据集上进行实验，结果表明MLAD达到或超过了现有主流模型性能，证明了其在LTSF任务中的有效性。

关键词: 长期时间序列预测（LTSF）, 多局部增强线性解码器（MLAD）, 局部特征增强机制（LFAM）

Abstract: In the field of long-term time series forecasting (LTSF), deep learning models with encoder-decoder architectures have demonstrated outstanding performance. Currently, although the encoder can extract deep-level temporal features from the input historical sequence, and capture the periodicity, trends and local information correlation within the time series, the decoder mostly relies on the global features output by the encoder, failing to fully utilize the local features extracted by the encoder, which limits the predictive performance of models. To fully explore and utilize local features, this paper proposes a multiple local augmented linear decoders (MLAD) approach. By introducing a local feature augmented mechanism (LFAM), it slides over the feature sequence generated by the encoder, and integrates the extracted local features with the original historical sequence to generate multiple local augmented sequences. The prediction result is determined by calculating the average of the prediction sequences corresponding to all local augmented sequences. Experiments conducted on 7 public datasets show that MLAD achieves or surpasses the performance of existing mainstream models, demonstrating its effectiveness in LTSF tasks.

Key words: long-term time series forecasting (LTSF), multiple local augmented linear decoders (MLAD), local feature augmented mechanism (LFAM)

陈子盎, 郭世伟, 马玉鹏, 韩云飞, 王保全. 面向长期时间序列预测的多局部增强线性解码器研究[J]. 计算机工程与应用, 2025, 61(16): 179-186.

CHEN Zi'ang, GUO Shiwei, MA Yupeng, HAN Yunfei, WANG Baoquan. Research on Multiple Local Augmented Linear Decoders for Long-Term Time Series Forecasting[J]. Computer Engineering and Applications, 2025, 61(16): 179-186.

参考文献

[1] 詹熙, 潘志松, 黎维, 等. 基于多尺度特征和注意力的金融时序预测方法[J]. 计算机工程与应用, 2022, 58(19): 107-115.
ZHAN X, PAN Z S, LI W, et al. Financial time series forecasting method based on multi-scale features and attention mechanism[J]. Computer Engineering and Applications, 2022, 58(19): 107-115.
[2] HEWAGE P, BEHERA A, TROVATI M, et al. Temporal convolutional neural (TCN) network for an effective weather forecasting using time-series data from the local weather station[J]. Soft Computing, 2020, 24(21): 16453-16482.
[3] NTI I K, TEIMEH M, NYARKO-BOATENG O, et al. Electricity load forecasting: a systematic review[J]. Journal of Electrical Systems and Information Technology, 2020, 7(1): 13.
[4] 高榕, 万以亮, 邵雄凯, 等. 面向改进的时空Transformer的交通流量预测模型[J]. 计算机工程与应用, 2023, 59(7): 250-260.
GAO R, WAN Y L, SHAO X K, et al. Traffic flow forecasting model for improved spatio-temporal transformer[J]. Computer Engineering and Applications, 2023, 59(7): 250-260.
[5] AHMED N K, ATIYA A F, EL GAYAR N, et al. An empirical comparison of machine learning models for time series forecasting[J]. Econometric Reviews, 2010, 29(5/6): 594-621.
[6] RICHMAN J S, MOORMAN J R. Physiological time-series analysis using approximate entropy and sample entropy[J]. American Journal of Physiology Heart and Circulatory Physiology, 2000, 278(6): H2039-H2049.
[7] GRAVES A. Long short-term memory[M]//Supervised sequence labelling with recurrent neural networks. Berlin, Heidelberg: Springer, 2012: 37-45.
[8] CHO K, VAN MERRIENBOER B, GULCEHRE C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[J]. arXiv:1406.1078, 2014.
[9] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems 30, 2017.
[10] NIU D X, YU M, SUN L J, et al. Short-term multi-energy load forecasting for integrated energy systems based on CNN-BiGRU optimized by attention mechanism[J]. Applied Energy, 2022, 313: 118801.
[11] 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.
[12] 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.
[13] ZHOU H Y, ZHANG S H, PENG J Q, et al. Informer: beyond efficient transformer for long sequence time-series forecasting[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(12): 11106-11115.
[14] NIE Y Q, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with transformers[J]. arXiv:2211.14730, 2022.
[15] WANG H, PENG J, HUANG F, et al. MICN: multi-scale local and global context modeling for long-term series forecasting[C]//Proceedings of the 11th International Conference on Learning Representations, 2023.
[16] LIU M H, ZENG A L, CHEN M X, et al. SCINet: time series modeling and forecasting with sample convolution and interaction[C]//Advances in Neural Information Processing Systems 35, 2022: 5816-5828.
[17] GONG Z Y, TANG Y J, LIANG J W. PatchMixer: a patch-mixing architecture for long-term time series forecasting[J]. arXiv:2310.00655, 2023.
[18] LIN S S, LIN W W, WU W T, et al. SegRNN: segment recurrent neural network for long-term time series forecasting[J]. arXiv:2308.11200, 2023.
[19] WU H X, HU T G, LIU Y, et al. TimesNet: temporal 2D-variation modeling for general time series analysis[C]//Proceedings of the 11th International Conference on Learning Representations, 2023.
[20] CHALLU C, OLIVARES K G, ORESHKIN B N, et al. NHITS: neural hierarchical interpolation for time series forecasting[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(6): 6989-6997.
[21] FAN W, ZHENG S, YI X H, et al. DEPTS: deep expansion learning for periodic time series forecasting[J]. arXiv:2203. 07681, 2022.
[22] OLIVARES K G, CHALLU C, MARCJASZ G, et al. Neural basis expansion analysis with exogenous variables: forecasting electricity prices with NBEATSx[J]. International Journal of Forecasting, 2023, 39(2): 884-900.
[23] DAS A, KONG W H, LEACH A, et al. Long-term forecasting with TiDE: time-series dense encoder[J]. arXiv:2304.08424, 2023.
[24] ZENG A L, CHEN M X, ZHANG L, et al. Are transformers effective for time series forecasting?[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(9): 11121-11128.
[25] LI Z, RAO Z W, PAN L J, et al. MTS-mixers: multivariate time series forecasting via factorized temporal and channel mixing[J]. arXiv: 2304. 08424, 2023.
[26] CHEN S A, LI C L, YODER N, et al. TSMixer: an all-MLP architecture for time series forecasting[J]. arXiv:2303.06053, 2023.
[27] ULYANOV D, VEDALDI A, LEMPITSKY V. Instance normalization: the missing ingredient for fast stylization[J]. arXiv:1607.08022, 2016.
[28] KIM T, KIM J, TAE Y, et al. Reversible instance normalization for accurate time-series forecasting against distribution shift[C]//Proceedings of the 10th International Conference on Learning Representations, 2022.