Ocean current speed prediction model in the sunda strait using Long Short-Term Memory (LSTM)

This research focuses on predicting the speed of ocean currents in the Sunda Strait by employing a Long Short-Term Memory (LSTM) model based on historical data. The approach includes data preprocessing, normalization of features using MinMaxScaler, segmentation of the data into training and testing sets, and the development of layered LSTM model architecture. The dataset comprises longitude, latitude, current velocity, and time information from 2022 to 2024. The findings indicate that the LSTM model can predict ocean current speeds with a Root Mean Squared Error (RMSE) of 13.66 cm/s, a mean absolute error (MAE) of 9.06 cm/s, and a determination coefficient (R²) of 0.87. The demonstration illustrated the typical design of ocean cunent speed fluctuations; however, forecasting unusual variations remains challenging. In summary, the LSTM model represents a practical approach for predicting ocean currents based on historical data, aiming to enhance prediction accuracy. This model will support navigation efforts and marine resource management in the Sunda Strait region.