A big-data-driven matching model based on deep reinforcement learning for cotton blending

Huosong Xia, Yuan Wang, Sajjad Jasimuddin, Justin Zuopeng Zhang, Andrew Thomas

Research output: Contribution to journalArticlepeer-review

1 Citation (SciVal)
40 Downloads (Pure)


China’s cotton textile industry is undergoing a critical period of digital transformation and upgrading to cope with pressure and challenges such as rising labour costs and large fluctuations in raw material prices. Developing a cost-based competitive advantage while ensuring a high-quality product is a critical problem in intelligent manufacturing. From the perspective of big data and reinforcement learning, the authors designed a reward value combining transaction, interaction, and measurement data by combining the reward mechanism and Markov decision for a combination of different raw materials in the intelligent textile factory. The authors propose a big data-driven application to the depth of the reinforcement learning to solve problems and build a big-data-driven matching model based on deep reinforcement learning to cotton matching. The offline strategy is designed to construct a memory bank and neural network, and the incentive mechanism of reinforcement learning is used to iterate the optimal yarn matching scheme to achieve the goal of intelligent cotton matching. The results show that deep reinforcement learning can be optimised using big data on the premise of quality assurance. Manufacturing costs can be optimised using a matching model of big data based on a deep reinforcement learning model.

Original languageEnglish
Pages (from-to)7573-7591
Number of pages19
JournalInternational Journal of Production Research
Issue number22
Early online date08 Dec 2022
Publication statusPublished - 2023


  • big data-driven
  • cotton blending cost optimisation
  • Deep reinforcement learning
  • matching model
  • reinforcement learning reward mechanism design


Dive into the research topics of 'A big-data-driven matching model based on deep reinforcement learning for cotton blending'. Together they form a unique fingerprint.

Cite this