LSTM and Bi-LSTM Based Prediction of Water Quality for Smart Aquaculture

Aquaculture is a significant factor in strengthening global food security and contributing to nutritional well-being. Water quality parameters including pH, salinity, dissolved oxygen (DO), and temperature significantly influence the health, survival, productivity, and sustainability of aquaculture. Deep Learning (DL) techniques for predicting aquaculture water quality (A-WQ) not only overcome the limitations inherent in traditional methods, but also pave the way for accurate and efficient outcomes. In this work, we propose the use of Long Short-Term Memory (LSTM) and Bi-directional Long Short-Term Memory (Bi-LSTM) Deep Learning-Recurrent Neural Network (DL-RNN) models to predict aquaculture water quality parameters. Moreover, this paper is presented with a comprehensive analysis of the effect of hyperparameters ( \(h_p\) ) on the prediction ability of the proposed Deep Learning models. The performance of the proposed LSTM and Bi-LSTM models is compared by means of both prediction accuracy and computational efficiency.