Deep Learning Approaches for Water Quality Prediction in Aquaponics Systems: A Comparative Study of Recurrent and Feedforward Architectures

Accurate prediction of water quality parameters is critical for the effective management and sustainability of aquaponics systems. This study evaluates the performance of four deep learning architectures: Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Simple Recurrent Neural Network (SimpleRNN), and Dense Neural Network (DenseNN) for forecasting key water quality parameters, including temperature, turbidity, dissolved oxygen, pH, ammonia, and nitrate. A significant research gap is addressed by analyzing how these models perform on noisy and minimally preprocessed datasets, advancing prior studies that lack robust preprocessing techniques tailored for aquaponics systems. A ten-fold cross-validation framework was employed to rigorously assess the models, with Mean Squared Error (MSE) and Mean Absolute Error (MAE) as evaluation metrics. The results demonstrate that LSTM and GRU models outperform other architectures, achieving average validation losses of 0.0028 and 0.0028, respectively, and mean absolute errors of 0.0473 and 0.0478. These models effectively capture the temporal dependencies inherent in time-series data, making them highly suitable for the complex dynamics of aquaponics systems. Unlike previous studies, this research highlights the trade-offs between computational efficiency and predictive accuracy in these models. In contrast, the SimpleRNN model exhibited higher error rates due to its inability to model long-term dependencies, while the DenseNN model, lacking temporal processing mechanisms, showed the lowest performance with an average validation loss of 0.0075 and MAE of 0.0797. This study underscores the importance of selecting appropriate model architectures for time-series forecasting tasks and provides a foundation for deploying predictive systems to optimize aquaponics operations. Future work includes exploring hybrid models with attention mechanisms and real-time data integration for enhanced operational efficiency.