TimeHF: Billion-Scale Time Series Models Guided by Human Feedback
This work addresses scalability and accuracy issues in time series forecasting for industrial applications like supply chain management, showing significant but incremental improvements over existing large time series model approaches.
The authors tackled the challenges of limited scalability, poor generalization, and suboptimal zero-shot performance in time series neural networks by introducing TimeHF, a pipeline for creating large time series models with 6 billion parameters that incorporates human feedback. Deployed in JD.com's supply chain for automated replenishment of over 20,000 products, it improved prediction accuracy by 33.21% over existing methods.
Time series neural networks perform exceptionally well in real-world applications but encounter challenges such as limited scalability, poor generalization, and suboptimal zero-shot performance. Inspired by large language models, there is interest in developing large time series models (LTM) to address these issues. However, current methods struggle with training complexity, adapting human feedback, and achieving high predictive accuracy. We introduce TimeHF, a novel pipeline for creating LTMs with 6 billion parameters, incorporating human feedback. We use patch convolutional embedding to capture long time series information and design a human feedback mechanism called time-series policy optimization. Deployed in JD.com's supply chain, TimeHF handles automated replenishment for over 20,000 products, improving prediction accuracy by 33.21% over existing methods. This work advances LTM technology and shows significant industrial benefits.