Explaining Salmon Harvest Deviation from Time Series

In Autumn 2024, ReLU focused on applying machine learning to better understand the operational practices at Grieg. We particularly focused on explaining the harvest deviation observed in their salmon farming process.

The project centered on model explainability, leveraging a time series dataset and combining modern machine learning techniques like XGBoost with Explainable AI (XAI) methods like SHAP. Additionally, we applied statistical learning approaches such as ElasticNet to further enhance interpretability.

Our final deliverable was a report presenting the key findings, including both domain-specific insights and theoretical takeaways. The delivery also included code implementations of the models and analyses conducted.


  • Grieg Seafood is an international seafood company specializing in fresh Atlantic salmon. Headquartered in Bergen, Norway, the company operates fish farms in Rogaland and Finnmark in Norway, British Columbia, and Shetland. Annually, Grieg Seafood produces significant quantities of salmon, primarily serving the North American and Asian markets.

    The company was founded as a subsidiary of the Grieg Group and is listed on the Oslo Stock Exchange. It has a history of expansion and innovation, including a merger with the Volden Group and past operations in Denmark and Chile. Grieg Seafood is recognized for its commitment to sustainable practices, having been awarded by the Global Aquaculture Alliance for Best Aquaculture Practices. It is a major player in the salmon farming industry, particularly in British Columbia, Canada.