Aquaculture operations generate a large number of variables and data. Data science methodologies can support informed decision making. However, given the emergence of this field, farmers are still not using these techniques to improve their decisions and knowledge about their production. The objective of this presentation is to illustrate the application of data analytic methodologies in a bream farm and in a seaweed farm in the context of the Valormar project (24517 supported by Compete2020, Lisboa2020, CRESC Algarve2020, PT2020 and the EU through FEDER/ERDF). Before carrying out the relevant data analytics a famer should define specific objectives and questions to be analysed. Also, a data flow for the analysis is needed, which requires a data management system for data gathering and assimilation. Datamining is an on-going task, during which are identified new questions, variables to be collected and further analysed
The first step to make available advance analytics to production managers is to jointly explore the data with visual datamining, in order to give visual tools for identification of drivers of growth, mortality, among other. As an example of the data exploration carried out at the fish farm, Figure 1 shows the influence of the start month of production in the fish growth per day and feed conversion ratio.
Machine learning algorithms were applied to create statistical models from historical data for prediction of key performance indicators (KPI) based on selected variables (predictors). Figure 2 illustrates the application of a Random Forest model to calculate the seaweed net harvest density of a given tank based on solar radiation, number of light hours, days in production and seeding density.