<![CDATA[Feeding is one of the key factors determining the success of catfish farming, as it directly affects fish growth, feed utilization efficiency, and the quality of the culture environment. This study aims to develop a Long Short-Term Memory (LSTM)-based feed requirement prediction model integrated into an Internet of Things (IoT)-based smart aquaponics system. The study adopts the CRISP-DM methodology, which consists of the stages of business understanding, data understanding, data preparation, modeling, evaluation, and implementation. Two models were developed: GrowthLSTM, which predicts the average fish weight, and PakanLSTM, which predicts feed requirements based on fish age, the number of fish, and the predicted fish weight. The evaluation results show that GrowthLSTM achieved an MAE of 2.3959 g, an RMSE of 3.4898 g, a MAPE of 3.19%, and an R² of 0.9962, while PakanLSTM achieved an MAE of 1.1034 g, an RMSE of 3.1447 g, a MAPE of 20.57%, and an R² of 0.9674. Both models were successfully integrated with an ESP32 through a REST API, enabling the system to perform automated feed prediction and dispensing to support more adaptive and efficient catfish farming.]]>
Copyrights © 2026
