Skip to search boxSkip to navigationSkip to main content

An adaptive method for fish growth prediction with empirical knowledge extraction

  • Hui Li
    ,
  • Yingyi Chen
    ,
  • Wensheng Li
    ,
  • Qingbin Wang
    ,
  • Yanqing Duan
    ,
  • Tao Chen
Research Output: Contribution to journal Article Peer-review

Open access

Abstract

Fish growth prediction provides important information for optimising production in aquaculture. Fish usually exhibit different growth characteristics due to the variations in the environment, the equipment used in different fish workshops and inconsistent application by operators of empirical rules varying from one pond to another. To address this challenge, the aim of this study is to develop an adaptive fish growth prediction method in response to feeding decision. Firstly, the practical operational experience in historical feeding decisions for different fish weights is extracted to establish the feeding decision model. Then, a fish weight prediction model is established by regression analysis methods based on historical fish production data analysis. The feeding decision model is integrated as the input information of the fish weight prediction model to obtain fish weight prediction. Furthermore, an adaptive fish growth prediction strategy is proposed by continuously updating model parameters using new measurements to adapt to specific characteristics. The proposed adaptive fish growth prediction method with empirical knowledge extraction is evaluated by the collected production data of spotted knifejaw (Oplegnathus punctatus). The results show that established models can achieve a good balance between goodness-of-fit and model complexity, and the adaptive prediction method can adapt to specific fish pond’s characteristics and provide a more effective way to increase fish weight prediction accuracy. The proposed method provides an important contribution to achieving adaptive fish growth prediction in a real time from the view of aquaculture practice for spotted knifejaw.

Publication Information

Output type

Research Output: Contribution to journal Article Peer-review

Original language

English

Pages from-to (Number of pages)

Pages 336-346 (11 pages)

Journal (Volume, Issue Number)

Biosystems Engineering (Volume 212)

Publication milestones

  • Accepted/In press - 10/11/2021
  • Published - 25/11/2021

Publication status

Published - 25/11/2021

ISSN

1537-5110

External Publication IDs

  • handle.net: 10547/625229
  • Scopus: 85120171274

Access to documents

Publication metrics

Metrics

Download statistics
Download count
20