Akademik Veri Yönetim Sistemi
Ocean Dynamics, cilt.76, sa.4, 2026 (SCI-Expanded, Scopus)
Because of the unpredictability of coastal features such as bathymetry and topography, and the uncertainty of driving factors and interactions among a variety of metocean elements, prediction of Sea Current Speed (SCS) in real-time is a complicated process. Although numerical approaches are extensively employed for this purpose, they require a lot of computational power and external data. In this study, in order to achieve one-hour ahead SCS forecasting, a long short-term memory deep learning approach (LSTM), an artificial intelligence technique such as an adaptive neuro-fuzzy inference system (ANFIS) with fuzzy c-means (FCM) and traditional statistical method, the Autoregressive Moving Average (ARMA) model were used. The performance criteria of the techniques are root-mean-square error (RMSE), mean absolute error (MAE), and correlation coefficient (R). Predicted values using ANFIS-FCM, LSTM, and ARMA methods were compared with the measured values by calculating their performance criteria values. The results from the comparison showed that the ARMA method provides the best results at 5 m, 15 m, and 25 m depths with the R-values of 0.9416, 0.9391, and 0.9339, respectively. In this context, it has been revealed that the ARMA technique outperforms other methods in terms of predictions at all depths and gives very sensitive results.