An Example of Artificial Intelligence and Its Use in the Aquaculture Industry

The paper presents a comparative analysis of artificial intelligence models applied to time series data obtained from three different aquaculture Facilities. The performance of the models was evaluated on a facility-by-facility basis; the data structure, the impact of environmental variables, and the role of time series characteristics on model success were highlighted. LSTM models have demonstrated high accuracy and explanatory power, particularly at Aquaculture Facilities A and B, due to their ability to learn long-term dependencies in time series data. This indicates that LSTM should be preferred at Aquaculture Facilities where the effects of environmental variables over time have a significant impact on fish growth rates. The Random Forest model has proven to be an effective alternative by achieving the highest explanatory power at Facility C, which has stable environmental conditions. Its interpretability and low computational cost provide advantages for field applications. MLP and SVM models have shown limited success in terms of data compatibility and parameter sensitivity. It is recommended that these models be trained with larger datasets and supported by hyperparameter optimization. In general, facility-based model selection, data preprocessing quality, and the accurate representation of environmental variables directly affect the success of artificial intelligence applications. Future studies could develop more robust decision support systems using hybrid model structures, attention mechanisms, and real-time data integration.