中国科学技术大学 · 认知智能全国重点实验室 University of Science and Technology of China · National Key Laboratory of Cognitive Intelligence

时间序列分析方向 Time Series Analysis

专注于时间序列分析的基础理论与前沿方法，覆盖典型任务、基础模型、分析智能体及领域应用，持续推动时序智能分析技术的研究边界。 Dedicated to foundational theory and cutting-edge methods for time series analysis, covering classic tasks, foundation models, analysis agents, and domain applications.

查看论文 Publications GitHub

最新动态 News

2026.02

观点论文《Beyond Model-Centric Prediction: Agentic Time Series Forecasting》发布于 arXiv。 Our position paper “Beyond Model-Centric Prediction: Agentic Time Series Forecasting” is now available on arXiv.

该工作系统讨论了时间序列预测从模型中心走向智能体中心的研究趋势，并提出新的问题设定与未来方向。 The paper outlines a shift from model-centric forecasting toward agentic workflows, highlighting problem settings and future directions.

arXiv
2026.01

两篇论文被 WSDM 2026 录用，分别聚焦慢思考 LLM 推理与时序自监督表示学习。 Two papers were accepted to WSDM 2026 on slow-thinking LLM reasoning and self-supervised time series representation learning.

相关成果包括 TimeReasoner 与 TimeMAE，持续推进时序分析与大模型结合的研究。 The accepted works include TimeReasoner and TimeMAE, extending our work on time series reasoning and foundation models.

Paper Code
2025.01

InstructTime 获评 Best of WSDM 2025，展示了多模态语言建模在时序分类中的潜力。 InstructTime was selected as Best of WSDM 2025, demonstrating the value of multimodal language modeling for time series classification.

该工作探索将时间序列分类重构为与自然语言和视觉语义协同建模的问题。 This work reformulates time series classification through joint modeling with language and visual semantics.

Paper Code

研究方向 Research Areas

01

时间序列分析基础理论 Foundations of Time Series Analysis

变点检测、谱分析、状态空间模型与统计检验等核心理论方法，构建时序分析的数学基础。 Core theoretical methods including change point detection, spectral analysis, state space models, and statistical testing.

02

时间序列典型任务 Classic Time Series Tasks

面向时间序列分类、异常检测与预测等核心任务，探索高效、鲁棒的深度学习建模方法。 Deep learning approaches for core tasks: classification, anomaly detection, and forecasting.

03

时序基础模型 Foundation Models for Time Series

自监督预训练与跨域迁移学习，构建可泛化的通用时序表示模型，赋能多任务下游应用。 Self-supervised pre-training and cross-domain transfer learning for universal, generalizable time series representations.

04

时序分析智能体 Time Series Analysis Agents

基于大语言模型的时序分析智能体，融合推理能力与领域知识，探索超越传统模型的新范式。 LLM-driven agentic frameworks integrating reasoning and domain knowledge for time series analysis beyond conventional models.

05

领域时序数据挖掘 Domain-specific Time Series Mining

电力负荷、能源功率、水文径流、云服务负载等重要领域时序数据的挖掘与预测研究。 Applied mining and forecasting for power load, energy generation, hydrological runoff, and cloud workload.

论文列表 Publications

Preprint

  • Preprint Agent

    Position: Beyond Model-Centric Prediction — Agentic Time Series Forecasting

    Mingyue Cheng, Xiaoyu Tao, Qi Liu, Ze Guo, Enhong Chen

    arXiv PDF
  • Preprint Survey

    A Comprehensive Survey of Time Series Forecasting: Concepts, Challenges, and Future Directions

    Mingyue Cheng, Zhiding Liu, Xiaoyu Tao, Qi Liu, Jintao Zhang, Tingyue Pan, Shilong Zhang, Panjing He, Xiaohan Zhang, Daoyu Wang, Jiahao Wang, Enhong Chen

    PDF GitHub
  • Preprint Classification

    InstructTime++: Time Series Classification with Multimodal Language Modeling via Implicit Feature Enhancement

    Mingyue Cheng, Xiaoyu Tao, Huajian Zhang, Qi Liu, Enhong Chen

    PDF Code
  • Preprint Forecasting

    AlphaCast: A Human Wisdom-LLM Intelligence Co-Reasoning Framework for Interactive Time Series Forecasting

    Xiaohan Zhang, Tian Gao, Mingyue Cheng*, Bokai Pan, Ze Guo, Yaguo Liu, Xiaoyu Tao

    PDF Code
  • Preprint Forecasting

    Time Series Forecasting as Reasoning: A Slow-Thinking Approach with Reinforced LLMs

    Yitong Zhou, Yucong Luo, Mingyue Cheng*, Jiahao Wang, Daoyu Wang, Tingyue Pan, Jintao Zhang, Qi Liu, Enhong Chen

    PDF Code
  • Preprint Forecasting

    MemCast: Memory-Driven Time Series Forecasting with Experience-Conditioned Reasoning

    Xiaoyu Tao, Mingyue Cheng*, Ze Guo, Shuo Yu, Yaguo Liu, Qi Liu, Shijin Wang

    PDF
  • Preprint Forecasting

    From Values to Tokens: An LLM-Driven Framework for Context-aware Time Series Forecasting via Symbolic Discretization

    Xiaoyu Tao, Shilong Zhang, Mingyue Cheng*, Daoyu Wang, Tingyue Pan, Bokai Pan, Changqing Zhang, Shijin Wang

    Code

2026

  • WSDM 2026 Forecasting

    Can Slow-Thinking LLMs Reason Over Time? Empirical Studies in Time Series Forecasting

    Mingyue Cheng, Jiahao Wang, Daoyu Wang, Xiaoyu Tao, Qi Liu*, Enhong Chen

    PDF Code
  • WSDM 2026 Foundation Anomaly

    TimeMAE: Self-Supervised Representations of Time Series with Decoupled Masked Autoencoders

    Mingyue Cheng, Xiaoyu Tao, Zhiding Liu, Qi Liu*, Hao Zhang, Rujiao Zhang, Enhong Chen

    PDF Code

2025

  • NeurIPS 2025 Forecasting

    Improving Time Series Forecasting via Instance-aware Post-hoc Revision

    Zhiding Liu, Mingyue Cheng, Guanhao Zhao, Jiqian Yang, Qi Liu, Enhong Chen

    PDF
  • ICML 2025 Foundation Anomaly

    TimeDART: A Diffusion Autoregressive Transformer for Self-supervised Time Series Representation

    Daoyu Wang, Mingyue Cheng*, Zhiding Liu, Qi Liu

    PDF Code
  • WSDM 2025 Classification Best of WSDM

    InstructTime: Advancing Time Series Classification with Multimodal Language Modeling

    Mingyue Cheng, Yiheng Chen, Qi Liu*, Zhiding Liu, Yucong Luo, Enhong Chen

    PDF Code
  • WSDM 2025 Foundation Anomaly

    Cross-Domain Pre-training with Language Models for Transferable Time Series Representations

    Mingyue Cheng, Xiaoyu Tao, Qi Liu*, Hao Zhang, Yiheng Chen, Defu Lian

    PDF Code
  • WWW 2025 Foundation Anomaly

    ConvTimeNet: A Deep Hierarchical Fully Convolutional Model for Multivariate Time Series Analysis

    Mingyue Cheng, Jiqian Yang, Tingyue Pan, Qi Liu*, Zhi Li

    PDF Code
  • CIKM 2025 Classification

    TableTime: Reformulating Time Series Classification as Zero-Shot Table Understanding via Large Language Models

    Jiahao Wang, Mingyue Cheng*, Qingyang Mao, Qi Liu, Feiyang Xu, Xin Li, Enhong Chen

    PDF Code
  • ACM TIST 2025 Classification

    Hierarchical Multimodal LLMs with Semantic Space Alignment for Enhanced Time Series Classification

    Xiaoyu Tao, Tingyue Pan, Mingyue Cheng*, Yucong Luo, Qi Liu, Enhong Chen

    PDF Code

2024

  • KDD 2024 Forecasting

    Generative Pretrained Hierarchical Transformer for Time Series Forecasting

    Zhiding Liu, Jiqian Yang, Mingyue Cheng*, Yucong Luo, Zhi Li

    PDF
  • NeurIPS 2024 Forecasting Anomaly

    Adaptive Normalization for Non-stationary Time Series Forecasting: A Temporal Slice Perspective

    Zhiding Liu, Mingyue Cheng, Zhi Li, Zhenya Huang, Qi Liu, Yanhu Xie, Enhong Chen

    Code

2023

  • WWW 2023 Classification

    FormerTime: Hierarchical Multi-scale Representation for Multivariate Time Series Classification

    Mingyue Cheng, Qi Liu*, Zhiding Liu, Zhi Li, Yucong Luo, Enhong Chen

    Code

开源项目 Open Source

Classification

InstructTime

面向时间序列分类的多模态语言建模框架，将时序信号与语言语义空间对齐。 A multimodal language modeling framework for time series classification with aligned signal and language semantics.

GitHub
Foundation

TimeMAE

基于解耦掩码自编码器的时序自监督表示学习项目，支持通用时序特征提取。 A self-supervised representation learning project built on decoupled masked autoencoders for general time series encoding.

GitHub
Forecasting

TimeReasoner

研究慢思考大语言模型在时间序列预测中推理能力的开源实现。 An open implementation for studying slow-thinking LLM reasoning in time series forecasting.

GitHub
Resource

Awesome Papers TSF

面向时间序列预测研究者整理的论文、方向与资源集合，便于快速了解领域进展。 A curated collection of papers, research directions, and resources for time series forecasting.

GitHub

经费支持 Funding

本方向的研究工作受到以下机构的支持，特此致谢。 Our research is generously supported by the following organizations.

国家自然科学基金委员会

National Natural Science Foundation of China

华为技术有限公司

Huawei Technologies Co., Ltd.