Akademik Veri Yönetim Sistemi
Tarim Bilimleri Dergisi, cilt.31, sa.1, ss.71-79, 2025 (SCI-Expanded)
Machine learning methods, which are one of the subfields of artificial intelligence and have gained popularity in applications in recent years, play an important role in solving many challenges in aquaculture. In this study, the relationship between changes in the physico-chemical characteristics of water and feed consumption was evaluated using machine learning methods. Eleven physico-chemical characteristics (temperature, pH, dissolved oxygen, electrical conductivity, salinity, nitrite nitrogen, nitrate nitrogen, ammonium nitrogen, total phosphorus, total suspended solids, and biological oxygen demand) of water were assessed. Among all the measured physico-chemical characteristics of water, temperature was determined to be the most important parameter to be evaluated in fish feeding. Moreover, pH2, EC2, TP2, TSS2, S2 and NO2-N parameters detected in the outlet water are more important than those detected in the inlet water in terms of feed consumption. Through regression analysis carried out using machine learning methods, the models developed with Random Forest, Gradient Boosting Machine and eXtreme Gradient Boosting algorithms exhibited higher success rates in predicting feed consumption compared to the other models. The present study highlights the pivotal role of machine learning methods in enhancing our understanding of fish feeding dynamics based on physico-chemical characteristics of water, thus contributing significantly to aquaculture management practices.