Sea level prediction using ARIMA, SVR and LSTM neural network: assessing the impact of ensemble Ocean-Atmospheric processes on models’ accuracy

This study aims to integrate a broad spectrum of ocean-atmospheric variables predict sea level variation along West Peninsular Malaysia coastline using machine learning and deep techniques. 4 scenarios different combinations such as surface temperature, salinity, density, atmospheric pressure, wind speed, total cloud cover, precipitation data were used train ARIMA, SVR LSTM neural network models. Results show that processes have more influence on prediction accuracy than ocean processes. Combining improves the model at all stations by 1- 9% for both LSTM. The means R optimal performing LSTM, ARIMA models are 0.853, 0.748 0.710, respectively. Comparison performance shows trained with is predicting except Pulua Langkawi where without performed best due dominating tide influence. suggests suitability vary across regions selecting an depends dominant physical governing variability in area investigation.