CoIFNet: A Unified Framework for Multivariate Time Series Forecasting with Missing Values

2025-06-16

AI Summary

Multivariate time series forecasting is often impaired by missing values caused by sensor failures, and existing two-stage “impute-then-forecast” paradigms suffer from objective misalignment and error accumulation. This paper proposes an end-to-end joint optimization framework that unifies imputation and forecasting objectives to eliminate error propagation. We introduce, for the first time, cross-timestep fusion (CTF) and cross-variable fusion (CVF) modules—rigorously proven to improve the theoretical upper bound on forecasting accuracy under missingness. The model leverages masked inputs and timestamp embeddings to ensure both robustness and computational efficiency. Extensive experiments demonstrate that, under 60% point-wise and block-wise missingness, our method reduces forecasting error by 24.40% and 23.81%, respectively, while reducing memory and computational overhead by 4.3× and 2.1× compared to state-of-the-art baselines.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Multi-modal Vision

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Multivariate time series forecasting (MTSF) is a critical task with broad applications in domains such as meteorology, transportation, and economics. Nevertheless, pervasive missing values caused by sensor failures or human errors significantly degrade forecasting accuracy. Prior efforts usually employ an impute-then-forecast paradigm, leading to suboptimal predictions due to error accumulation and misaligned objectives between the two stages. To address this challenge, we propose the Collaborative Imputation-Forecasting Network (CoIFNet), a novel framework that unifies imputation and forecasting to achieve robust MTSF in the presence of missing values. Specifically, CoIFNet takes the observed values, mask matrix and timestamp embeddings as input, processing them sequentially through the Cross-Timestep Fusion (CTF) and Cross-Variate Fusion (CVF) modules to capture temporal dependencies that are robust to missing values. We provide theoretical justifications on how our CoIFNet learning objective improves the performance bound of MTSF with missing values. Through extensive experiments on challenging MSTF benchmarks, we demonstrate the effectiveness and computational efficiency of our proposed approach across diverse missing-data scenarios, e.g., CoIFNet outperforms the state-of-the-art method by $underline{ extbf{24.40}}$% ($underline{ extbf{23.81}}$%) at a point (block) missing rate of 0.6, while improving memory and time efficiency by $underline{oldsymbol{4.3 imes}}$ and $underline{oldsymbol{2.1 imes}}$, respectively.

Problem

• Addresses multivariate time series forecasting with missing values
• Unifies imputation and forecasting to improve accuracy
• Enhances robustness to missing data in diverse scenarios

Innovation

• Unifies imputation and forecasting in one framework
• Uses Cross-Timestep and Cross-Variate Fusion modules
• Improves performance bound with theoretical justification