One step ahead - to predict fish growth and fitness

Scientists develop a web-based solution that enables fish farmers to be one step ahead – to predict biomass, fish growth, and health status in farmed salmon.

Sist oppdatert: Aug 19, 2019
Published Apr 26, 2019
Alla Sapronova Portrett Trappa Web
Alla Sapronova. (Photo: Andreas R. Graven)

A new research project aims to take the fish farming industry’s sustainability and profitability to a new level - with help from machine learning and Big Data.

The 2-year project named Predict-fit: intelligent predictive tool for value creation in aquaculture, is owned by VIS. NORCE is the scientific partner, and the project - which will run till December 2020 - is funded by The Research Council of Norway

– This project has the unique combination of a proactive concept including both machine learning and fish biology competence in the forefront of research, says data scientist and project leader Alla Sapronova at NORCE.

Her expertise is in both artificial intelligence, image recognition and machine learning. Sapronova develops a new model for the project:

– I train computers in the same way one teach children. I show the computer patterns of input signals and tell it what I expect the output signal to be. I repeat this process until the system begins to recognize the patterns. Then I show the computer an input signal, that it has not seen before and test whether the system understands what it is, Sapronova explains.

The fish farming industry needs reliable and accurate forecast of fish growth - for both weight and length - that will increase operational planning accuracy and profits for the companies.

– Apart from this, it is important for fish farmers to have better control of fish health evaluation and prognosis, to support proactive measures and reduce loss of biomass, Sapronova underscores.

To develop the new model, Sapronova uses environmental data, food data and fish biology data. The fish farming company Grieg Seafood has provided data to the project: time series from several fish tanks at one fish farm location.

The working method for the model is complex. Here the approach is explained very briefly:

– Developing the data model has to do with supervised and unsupervised learning and pattern search in all data we have available. One can call it a holistic approach, in the sense that all data is taken in account to supply the end user - the fish farming companies - with a decision support system, Sapronova says.

The NORCE data scientist underscores that they aim to make a proactive solution for the fish farmers.

– This means basically optimization of production and being one step ahead. If something is not going the optimal way, the web based solution can tell which parameters to change, for example temperature or feed, Sapronova says.

Partners in the project are the fish farming company Kvarøy fiskeoppdrett, and AKVA-Group.

– We are working with different kinds of partners, and more from the Bergen area would be beneficial.