Health prediction for king salmon via evolutionary machine learning with genetic programming.

Area of Science:

• Aquaculture
• Machine Learning
• Animal Health

Background:

• King (Chinook) salmon farming is significant in Aotearoa New Zealand, representing over half of global production.
• Accurate health assessment of farmed king salmon is crucial but challenging due to complex environmental and biological factors.
• Evolutionary machine learning (EML) shows promise for complex prediction tasks but hasn't been applied to king salmon health.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting king salmon health in Aotearoa New Zealand.
• To investigate the efficacy of genetic programming (GP), an EML algorithm, for this specific application.
• To assess GP's ability to identify key health indicators and generate interpretable predictive models.

Main Methods:

• Data processing techniques were applied to prepare datasets for king salmon health prediction.
• A novel health prediction method utilizing genetic programming was designed and implemented.
• The performance of GP was compared against other standard machine learning algorithms.

Main Results:

• Genetic programming demonstrated superior overall performance in most trials compared to other machine learning algorithms.
• GP effectively identified important features for classifying king salmon health status.
• The models developed using GP were found to be potentially interpretable.

Conclusions:

• Genetic programming is a highly effective tool for king salmon health prediction in aquaculture.
• This research provides a significant advancement in developing automated health assessment tools for farmed king salmon.
• The findings support the adoption of GP for enhancing health management in New Zealand's king salmon industry.