Ship Trajectory Prediction Based On TSO-LSTM

With the rapid development of the shipbuilding industry,water traffic problems are becoming increasingly prominent,and ship navigation trajectory prediction has become an important technical means.In vessel traffic service,the navigation track data is an important reference information.By utilizing the existing historical navigation trajectory data of ships,predict the future navigation trajectory of ships.This article will revolve around the theme of "Ship Navigation Trajectory Prediction",based on the theoretical foundations of clustering algorithms and deep learning,using AIS data of ship navigation and Python language to complete the implementation of clustering algorithms and neural network prediction models.Through a combination of theoretical research and practical experiments,research on ship navigation trajectory prediction will be carried out,mainly including the following aspects of work:Due to a series of issues such as communication,waves,and water areas during ship navigation,the raw AIS trajectory data obtained may have missing AIS data within certain time intervals.Interpolation algorithms are used to fill in the missing data and ensure the continuity and integrity of the ship’s trajectory data.By comparing various interpolation algorithms and selecting the most suitable one,the missing AIS data can be effectively filled.To address the issue of large water areas in ship trajectory distribution,the K-Means clustering algorithm is utilized to cluster the interpolated ship trajectory data.By dividing the trajectory data based on water areas,the trajectory points within the same water area are clustered together to form clusters.Each cluster is labeled,and the clustered and labeled trajectory data are visually displayed.This allows for a more intuitive observation of the distribution of trajectory data in different water areas,facilitating subsequent analysis and processing.To address the issue of noise points in ship trajectory data caused by communication,weather conditions,anchoring,etc.,the DBSCAN clustering algorithm is employed to denoise the ship trajectory data after water area partitioning.By identifying and removing noise points in the ship’s trajectory,more accurate ship trajectory clusters can be obtained,enabling more precise ship trajectory prediction.To address the issues of slow convergence speed and low prediction accuracy in LSTM models,a solution is proposed by combining the Tuna Swarm Optimization(TSO)algorithm.The TSO algorithm is used to optimize the hyperparameters of the LSTM network model,thereby improving the convergence and prediction performance of the model.A comparison is made between TSO-LSTM and traditional LSTM models,and the predicted results of both models are compared with the actual ship trajectory data.The results demonstrate that the ship trajectory prediction model based on TSO-LSTM exhibits higher prediction accuracy and robustness,thereby providing assistance in maritime traffic safety management.