| This study investigates the application of Long Short-Term Memory (LSTM) networks for predicting marine environmental variables, focusing on the challenges of analyzing time-dependent marine data. Our research applies LSTM models to predict both single and multiple marine environmental factors with remarkable precision. In the single-variable scenario, the model achieved a Mean Absolute Error (MAE) of 5.43E-09 in training and 8.10E-09 in testing, with a Root Mean Squared Error (RMSE) of 8.99E-09 in training and 1.44E-08 in testing. For the multiple-variable scenario, the model demonstrated an MAE of 2.40E-01 in training and 2.26E-01 in testing, alongside an RMSE of 3.67E-01 in training and 3.02E-01 in testing. These results illustrate the LSTM model’s capability to handle the complexity of marine data accurately, marking a significant advance in marine environmental prediction accuracy and reliability. The findings suggest promising avenues for model precision enhancement and validation across varied marine environments, offering new prospects for marine ecosystem management and disaster prevention strategies. |
