计算机科学

计算机科学 ›› 2025, Vol. 52 ›› Issue (2): 91-98.doi: 10.11896/jsjkx.240400127

• 数据库&大数据&数据科学 • 上一篇    下一篇

基于多粒度多尺度深度时空模型的长期序列预测方法

陈加毫1, 谢良1, 廖思灏1, 吴雨琛1, 徐海蛟2   

  1. 1 武汉理工大学理学院 武汉 430070
    2 广东第二师范学院计算机学院 广州 510303
  • 收稿日期:2024-04-17 修回日期:2024-10-20 出版日期:2025-02-15 发布日期:2025-02-17
  • 通讯作者: 谢良(whutxl@hotmail.com)
  • 作者简介:(2248919270@qq.com)
  • 基金资助:
    广东省自然科学基金(2020A1515011208);广州市基础研究教计划基础与应用基础研究项目(202102080353);广东省普通高校自然科学类特色创新项目(2019KTSCX117)

Long-term Series Forecasting Method Based on Multi-granularity Multi-scale Deep Spatio-TemporalModeling

CHEN Jiahao1, XIE Liang1, LIAO Sihao1, WU Yuchen1, XU Haijiao2   

  1. 1 College of Science,Wuhan University of Technology,Wuhan 430070,China
    2 School of Computer Science,Guangdong University of Education,Guangzhou 510303,China
  • Received:2024-04-17 Revised:2024-10-20 Online:2025-02-15 Published:2025-02-17
  • About author:CHEN Jiahao,born in 1997.His main research interests include deep learning and data mining.
    XIE Liang,born in 1987,Ph.D,asso-ciate prefessor.His main research intere-sts include multimedia retrieval and machine learning.
  • Supported by:
    Natural Science Foundation of Guangdong Province,China(2020A1515011208),Basic and Applied Basic Research Project of Guangzhou Basic Research Teaching Program(202102080353) and Characteristic Innovation Project of Natural Science in General Colleges and Universities in Guangdong Province(2019KTSCX117).

PDF (PC)

164

摘要： 时间序列建模一直是金融和交通等多个领域研究的热点,时空模型因能更全面捕捉时序数据的复杂关联和趋势,受到研究者们的广泛关注。近年来,基于时空模型的长期序列预测取得显著成果,但现有方法受到多粒度或多尺度研究的限制,无法充分挖掘数据的时空信息。为解决这一问题,提出了一种多粒度多尺度深度时空模型(MMDSTM)。该模型首先通过分解初始数据获取季节、周期和粒度序列;然后,利用基于多尺度等距卷积生成尺度预测,利用基于注意力的时空特征层生成多粒度预测;最后,通过多层次融合合并多粒度与多尺度预测的预测结果。在实验中,MMDSTM相比近期的新方法在股票、交通和电池数据集上MSE指标分别下降了6.2%,21.5%和1%。多粒度和多尺度的引入显著提升了时间序列预测性能。

关键词: 多粒度学习, 多尺度学习, 时间序列预测, 金融市场, 交通流速度

Abstract: Time series modeling has been the focus of research in a number of fields,including finance and transportation,and spatio-temporal models have received a lot of attention from researchers because of their ability to capture the complex associations and trends in time-series data more comprehensively.In recent years,long-term series forecasting based on spatio-temporal modeling has achieved remarkable results,but the existing methods are limited by multi-granularity or multi-scale studies,which cannot fully mine the spatio-temporal information of the data.To overcome this problem,a multi-granularity multi-scale deep spatio-temporal model(MMDSTM) is proposed.The model first obtains seasonal,periodic and granularity sequences by decomposing the initial data.Then,the multi-scale isometric convolution generates scale predictions,while attention-based spatio-temporal feature layers generates multi-granularity predictions.Finally,the prediction results of multi-granularity and multi-scale predictions are merged by multi-level fusion.In experiments,MMDSTM's MSE metric decreases by 6.2%,21.5% and 1% compared to other methods on stock,traffic and battery datasets,and the introduction of multi-granularity and multi-scale significantly improves the time series forecasting performance.

Key words: Multi-granularity learning, Multi-scale learning, Time series forecasting, Financial markets, Traffic flow speed

中图分类号: 

  • TP391
[1]SALINAS D,FLUNKERT V,GASTHAUS J,et al.DeepAR:Probabilistic forecasting with autoregressive recurrent networks[J].International Journal of Forecasting,2020,36(3):1181-1191.
[2]BAHDANAU D,CHO K,BENGIO Y.Neural machine transla-tion by jointly learning to align and translate[J].arXiv:1409.0473,2014.
[3]BELGHZI M I,BARATIN A,RAJESHWAR S,et al.Mutual information neural estimation[C]//International Conference on Machine Learning.PMLR,2018:531-540.
[4]BINKOWSKI M,MARTI G,DONNAT P.Autoregressive con-volutional neural networks for asynchronous time series[C]//International Conference on Machine Learning.PMLR,2018:580-589.
[5]BOX G.BOX and JENKINS:time series analysis,forecastingand control[M]//A Very British Affair:Six Britons and the Development of Time Series Analysis During the 20th Century.London:Palgrave Macmillan UK,2013:161-215.
[6]VAN GESTEL T,SUYKENS J A K,BAESTAENS D E,et al.Financial time series prediction using least squares support vector machines within the evidence framework[J].IEEE Transactions on Neural Networks,2001,12(4):809-821.
[7]BOX G E P,PIERCE D A.Distribution of residual autocorrela-tions in autoregressive-integrated moving average time series models[J].Journal of the American statistical Association,1970,65(332):1509-1526.
[8]YOO J,SOUN Y,PARK Y,et al.Accurate multivariate stockmovement prediction via data-axis transformer with multi-level contexts[C]//Proceedings of the 27th ACM SIGKDD Confe-rence on Knowledge Discovery & Data Mining.2021:2037-2045.
[9]DING Q,WU S,SUN H,et al.Hierarchical Multi-Scale Gaus-sian Transformer for Stock Movement Prediction[C]//IJCAI.2020:4640-4646.
[10]HAN M,XU M.Laplacian echo state network for multivariate time series prediction[J].IEEE Transactions on Neural Networks and Learning Systems,2017,29(1):238-244.
[11]HOCHREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[12]VARGAS M R,DE LIMA B S L P,EVSUKOFFA G.Deep learning for stock market prediction from financial news articles[C]//2017 IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications(CIVEMSA).IEEE,2017:60-65.
[13]NIEPERT M,AHMED M,KUTZKOV K.Learning convolu-tional neural networks for graphs[C]//International Conference on Machine Learning.PMLR,2016:2014-2023.
[14]CHEN W,JIANG M,ZHANG W G,et al.A novel graph convolutional feature based convolutional neural network for stock trend prediction[J].Information Sciences,2021,556:67-94.
[15]YU B,YIN H,ZHU Z.Spatio-temporal graph convolutional networks:A deep learning framework for traffic forecasting[J].arXiv:1709.04875,2017.
[16]LAI G,CHANG W C,YANG Y,et al.Modeling long-and short-term temporal patterns with deep neural networks[C]//The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval.2018:95-104.
[17]SHIN S Y,SUN F K,LEE H.Temporal pattern attention for multivariate time series forecasting[J].Machine Learning,2019,108:1421-1441.
[18]VASWANI A,SHAZEERN,PARMAR N,et al.Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.2017:6000-6010.
[19]WU Z,PAN S,LONG G,et al.Connecting the dots:Multiva-riate time series forecasting with graph neural networks[C]//Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining.2020:753-763.
[20]HOU M,XU C,LIU Y,et al.Stock trend prediction with multi-granularity data:A contrastive learning approach with adaptive fusion[C]//Proceedings of the 30th ACM International Confe-rence on Information & Knowledge Management.2021:700-709.
[21]HOU M,XU C,LI Z,et al.Multi-Granularity Residual Learning with Confidence Estimation for Time Series Prediction[C]//Proceedings of the ACM Web Conference 2022.2022:112-121.
[22]WANG M,WANG H,ZHANG F.FAMC-Net:Frequency Domain Parity Correction Attention and Multi-Scale Dilated Convolution for Time Series Forecasting[C]//Proceedings of the 32nd ACM International Conference on Information and Know-ledge Management.2023:2554-2563.
[23]WANG H,PENG J,HUANG F,et al.Micn:Multi-scale local and global context modeling for long-term series forecasting[C]//The Eleventh International Conference on Learning Re-presentations.2022.
[24]JOUIN M,GOURIVEAU R,HISSEL D,et al.Prognostics ofPEM fuel cell in a particle filtering framework[J].International Journal of Hydrogen Energy,2014,39(1):481-494.
[25]CHUNG H,SHIN K.Genetic algorithm-optimized multi-channel convolutional neural network for stock market prediction[J].Neural Computing and Applications,2020,32(12):7897-7914.
[26]FAN J,SHEN Y.StockMixer:A Simple Yet Strong MLP-Based Architecture for Stock Price Forecasting[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2024:8389-8397.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市渝北区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发