Laboratoire de Génie Informatique et d’Automatique de l’Artois

The rapid growth of the global population, urbanization, and increasing demand for protein-rich diets have intensified pressure on global food supply chains. Aquatic food has become a key component of human nutrition and one of the most traded food commodities worldwide. However, the sustainability of seafood production systems is increasingly challenged by resource overexploitation and environmental constraints, raising concerns about long-term food security.

Forecasting seafood demand is particularly complex due to the non-linear, seasonal, and dynamic nature of consumption patterns. These patterns are influenced by environmental and climatic factors, market dynamics, and regional characteristics. Moreover, available data are often fragmented, noisy, and incomplete, which limits the effectiveness of conventional forecasting approaches.

This thesis aims to develop a smart and flexible forecasting framework capable of learning from diverse and evolving seafood demand patterns. The proposed data-driven approach focuses on adaptability, robustness to uncertainty, and the ability to handle complex temporal structures across heterogeneous data sources. The objective is to improve forecasting accuracy and reliability, thereby supporting informed decision-making for the sustainable planning and management of seafood production systems.