Background: Electrical conductivity (EC) is an important indicator of surface water quality, primarily influenced by temperature, salinity, and human activities. The conventional EC experimental technique is resource-intensive and time-consuming. Recent advancements in machine learning provide an innovative technique for accurate EC prediction using historical time series data.
Methods: Surface water EC was assessed via four machine learning techniques, namely Artificial Neural Network (ANN), Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM). The predictive capability of the aforementioned models was assessed via
six statistical performance indicators, namely Coefficient of determination (R2), Percent Bia (PBias), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Relative Root Mean Square Error (RRMSE), and Nash-Sutcliffe Efficiency (NSE).
Results: The findings of the present research work show that the LSTM model outperforms in predicting EC. The LSTM model’s efficacy was demonstrated by its outstanding R2 values of 0.99 and 0.94 during training and testing, respectively. Notably, RNN, ANN, and CNN ranked second, third, and fourth, respectively, based on statistical performance indicators.
Conclusion: The results show that LSTM outperforms the remaining models in predicting EC. The findings of this study can assist water quality managers in finding the optimum machine learning model for modeling EC in the understudied area. Overall, this work advances our understanding of EC prediction using machine learning techniques.