DEVELOPMENT OF MACHINE LEARNING COMPONENTS FOR WATER QUALITY PREDICTION IN RED TILAPIA FARMING

• Main Author: Marcelino, Arjuna
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/76034
• Institution: Institut Teknologi Bandung

Water quality monitoring and prediction are essential for red tilapia farming. Water quality dramatically affects the health and growth of fish. Using machine learning to predict water quality can help farmers take appropriate actions to maintain optimal conditions for fish growth. This study aims to obtain a suitable algorithm (LSTM or GRU) to develop a model that performs better in the context of water quality prediction in red tilapia farming. This study mainly compared algorithms to see which algorithm gave more accurate prediction results. Water quality data were collected from a monitoring system installed in red tilapia farming ponds. The data is pre-processed before being used in learning. After that, the model is implemented and trained using that data. Experiments were conducted on both algorithms for obtaining architecture, preprocessing techniques, window length, and the best hyperparameters. Model performance was evaluated using MAPE, RMSE, and MAE evaluation metrics. This study found that the LSTM model provides more accurate water quality prediction results than the GRU model. The best model was tested with different red tilapia ponds. Testing provided accurate predictive results. The LSTM was used as a machine learning component for water quality prediction in water quality management systems in red tilapia farming.