Bridging the Last Mile of Prediction: Enhancing Time Series Forecasting with Conditional Guided Flow Matching

2025-07-09

AI Summary

Diffusion models for time-series forecasting suffer from fixed source distributions and inflexible, unidirectional sampling paths; while flow matching offers computational efficiency and modeling flexibility, its potential to explicitly model auxiliary model prediction errors remains unexplored. This paper proposes Conditional-Guided Flow Matching (CGFM), the first method to incorporate prior model prediction errors as explicit conditioning signals—jointly with historical observations—to construct a bidirectional conditional probability path. To enhance expressivity and robustness, CGFM introduces an affine transformation to expand the path space. Integrating flow matching, error-guided conditioning, and a general affine path design, CGFM balances high-fidelity generation with inference efficiency. Extensive experiments across diverse time-series forecasting benchmarks demonstrate that CGFM consistently outperforms state-of-the-art methods, achieving superior accuracy, strong generalization across domains, and practical effectiveness.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Reasoning under Uncertainty: Relational Probabilistic Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Diffusion models, a type of generative model, have shown promise in time series forecasting. But they face limitations like rigid source distributions and limited sampling paths, which hinder their performance. Flow matching offers faster generation, higher-quality outputs, and greater flexibility, while also possessing the ability to utilize valuable information from the prediction errors of prior models, which were previously inaccessible yet critically important. To address these challenges and fully unlock the untapped potential of flow matching, we propose Conditional Guided Flow Matching (CGFM). CGFM extends flow matching by incorporating the outputs of an auxiliary model, enabling a previously unattainable capability in the field: learning from the errors of the auxiliary model. For time series forecasting tasks, it integrates historical data as conditions and guidance, constructs two-sided conditional probability paths, and uses a general affine path to expand the space of probability paths, ultimately leading to improved predictions. Extensive experiments show that CGFM consistently enhances and outperforms state-of-the-art models, highlighting its effectiveness in advancing forecasting methods.

Problem

• Enhancing time series forecasting with flexible flow matching
• Overcoming rigid source distributions in diffusion models
• Learning from prior model errors for better predictions

Innovation

• Conditional Guided Flow Matching enhances forecasting
• Integrates auxiliary model errors for learning
• Uses affine paths to improve predictions