Temporal-spatial dependence-based method for ship traffic flow prediction in wind farms water area

To enhance the precision of predicting ship traffic flow in complex maritime areas, this research examines various traffic flow segments by considering the traffic flow structure and develops a Gated Recurrent Unit (GRU) neural network model. Instead of using the conventional time-series input, the study introduces a spatio-temporal dependency feature matrix for ship traffic flow forecasting. The proposed model is implemented to forecast traffic flow in the wind farm cluster waters of Yancheng, Jiangsu province, and is compared against benchmark models including Autoregressive Integrated Moving Average (ARIMA), Support Vector Machines (SVM), and Long Short-Term Memory (LSTM). The findings indicate that the GRU model, utilizing spatio-temporal dependencies, outperforms mainstream models in accuracy, confirming its effectiveness and dependability.