Model Prediksi Kualitas Air Untuk Budidaya Ikan Lele Dengan Algoritma Extreme Gradient Boosting

The growth of the aquaculture sector in Indonesia, particularly in catfish farming, has experienced significant increases. However, a major challenge faced is the need for accurate predictions of fish development to optimize production and minimize the risk of losses. This study aims to develop a growth prediction system for catfish based on machine learning using the XGBoost algorithm, which considers critical environmental factors such as water quality (temperature, pH, dissolved oxygen, ammonia, and nitrate). With this system, catfish farmers can monitor water quality in real-time, allowing them to take timely and optimal preventive actions regarding feed provision, thereby improving harvest yields and reducing operational costs. The XGBoost model demonstrates good performance with a Mean Absolute Error (MAE) of 0.073 for fish weight and 14.66 for fish length, a Mean Squared Error (MSE) of 0.123 for fish weight and 1.278 for fish length, and an R² value of 0.998 for both variables, indicating high accuracy in predicting fish growth. It is expected that this research will not only enhance productivity and efficiency in catfish farming but also support digital transformation in Indonesia's fisheries sector, providing a competitive advantage for farmers in facing increasingly complex industry challenges.