Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Qi, Yongzhi, Hu, Hao, Lei, Dazhou, Zhang, Jianshen, Shi, Zhengxin, Huang, Yulin, Chen, Zhengyu, Lin, Xiaoming, Shen, Zuo-Jun Max
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2501.15942
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913666918514688
author Qi, Yongzhi
Hu, Hao
Lei, Dazhou
Zhang, Jianshen
Shi, Zhengxin
Huang, Yulin
Chen, Zhengyu
Lin, Xiaoming
Shen, Zuo-Jun Max
author_facet Qi, Yongzhi
Hu, Hao
Lei, Dazhou
Zhang, Jianshen
Shi, Zhengxin
Huang, Yulin
Chen, Zhengyu
Lin, Xiaoming
Shen, Zuo-Jun Max
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2501_15942
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TimeHF: Billion-Scale Time Series Models Guided by Human Feedback
Qi, Yongzhi
Hu, Hao
Lei, Dazhou
Zhang, Jianshen
Shi, Zhengxin
Huang, Yulin
Chen, Zhengyu
Lin, Xiaoming
Shen, Zuo-Jun Max
Machine Learning
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.
title TimeHF: Billion-Scale Time Series Models Guided by Human Feedback
topic Machine Learning
url https://arxiv.org/abs/2501.15942