A Marine Traffic Pattern Prediction System Based on Graph Convolutional and Kolmogorov-Arnold Networks

In recent days, Marine Traffic has been an analytic provider that tracks the movement of ships and provides data like the ship's location, ship details, and weather conditions. Moreover, marine traffic supports navigators and improves the maritime traffic management system. Marine traffic can be a major cause of the collision among the ships, and it also causes the missing of the routes. To, overcome these issues this paper proposed a Kolmogorov-Arnold Networks (KAN) algorithm for prediction, effectively achieving better accuracy. Initially, data obtained from the Automatic Identification System (AIS) dataset is given as input for the preprocessing using the Pearson Correlation Coefficient (PCC) which is employed to analyze the traffic flow in the port. Later, the Density-Based Spatial Clustering of Applications with Noise (DBSAN) clustering is utilized to discover where vertex points represent a waypoint. Then Graph Convolutional Networks (GCNs) are used for the trajectory prediction to process inland waterway traffic network signals. Finally, KAN incorporates learnable activation functions that take historical weather data points as inputs, and predict resilience indices as outputs. The performance of the proposed method is evaluated with various performance metrics such as MAE, MSE, RMSE, and MAPE metrics on the AIS dataset.