GHPSO-ATLSTM: a novel attention-based genetic LSTM to predict water quality indicators

Water demand and water quality are the two important aspects that can never be compromised at any point. In recent years, urbanization and industrialization have significantly degraded the quality of accessible surface water. Improving the monitoring capabilities of surface water may result in accurate prediction, which can assist policymakers in developing a strategy to reduce water pollution. Traditional monitoring techniques are time-consuming and expensive, making it difficult to satisfy the demands of real-time visualization in current situations. To deal with this challenge, a novel approach (GHPSO-ATLSTM) is developed to predict COD in surface water. The optimal features are selected using a genetic algorithm (GA), and the hyperparameters of LSTM are optimized with hidden particle swarm optimization (HPSO) followed by an attention (AT) layer to enhance the prediction accuracy. The proposed method is compared to the GRU, RNN, and MLR models, and performance metrics are used to pick the best algorithm for prediction. For the proposed method, \(R^2\) $R^2$ ranges between 0.91 to 0.96, MAE lies between 0.53 to 0.9 mg/l, MSE remains between 0.4 to 1.2 mg/l and MAPE ranges between 6.9 to 11%. The outcome reveals that the suggested model performs better and predicts the COD more accurately with respect to other models. The suggested strategy works efficiently whenever we know the target variable and is scalable across all domains. Prediction of water quality features is crucial for sustainable water resources management as it allows for proper decision-making and optimal resource allocation to maintain and enhance water quality.