Sea Level Time Series Forecasting by using LSTM with Attention Mechanism, A Case Study in Jakarta, Indonesia

Forecasting sea level is crucial, especially in areas located at the interface between land and ocean, as it can facilitate early and appropriate actions to reduce environmental, social, and economic damages caused by high sea levels. Sea level forecasting is used to estimate the accuracy in successfully addressing emergency situations due to floods. This research discusses a new approach to forecasting sea level using Long Short-Term Memory (LSTM) equipped with Attention mechanisms. The research focuses on the application of this method in a case study in Jakarta, Indonesian, with the aim of improving the accuracy and reliability of sea level forecasting. The importance of sea level predictions lies in their ability to contribute to flood risk mitigation in the coastal areas of Jakarta. By using the LSTM with Attention method, this research aims to address the complexity of patterns in sea level time series data, so that the prediction results can serve as a more reliable basis for disaster risk management decisions. The LSTM with Attention method is designed to model complex patterns in sea level time series data. Experiments were conducted using sea level datasets in Jakarta obtained through the Sea Level Station Monitoring Facility (SLSMF). The analysis process is divided into two parts: training data and testing data. This research compares the performance of the LSTM with Attention method with two other methods, namely RNN and LSTM, to evaluate prediction accuracy over a 3-month period from June to August 2023. The results show that the LSTM method outperforms RNN and LSTM with Attention in forecasting sea levels. Qualitative and quantitative evaluations using R2 and RMSE confirm the superiority of LSTM. This research also involves long-term predictions for periods of 3, 5, 7, 10, and 14 days ahead, providing further insights into the model's ability to anticipate changes in sea level.