Predicting spatio-temporal dynamics in aquaculture networks: An extended Katz index approach.

Abstract

The effective surveillance of the distribution of live fish between aquaculture farms is crucial for maintaining food security and preventing disease outbreaks. However, existing conventional models often assume the network is static and do not incorporate other factors that contribute to movement between farms, lacking the ability to accurately predict future movements, especially given the dynamic interactions within aquaculture networks. This study addresses this gap by developing the Edge-Weighted Katz Index (EWKI), an extension of the traditional Katz index that integrates spatial information to improve the accuracy of predicting fish distribution between farms. Using a comprehensive dataset on the distribution of live fish between farms in England and Wales from the year 2010 and 2023, the study evaluates the performance of the EWKI model in comparison to other similarity-based link prediction methods. The results indicate that the EWKI model significantly outperforms other methods, achieving a precision of 92.89%, a recall of 81.09%, and an F1-score of 86.59%, alongside an AUPR of 93.44% and an AUROC of 99.97%. This research has practical implications, as the developed method can accurately predict the distribution of fish between farms, supporting predictions of disease spread and facilitating targeted interventions. Furthermore, the integration of spatial information into the network analysis has broader applications across various fields where understanding and predicting spatially influenced network dynamics are crucial, including transportation networks.