| In recent years,with the increasing proportion of shipping in international trade and transportation,there are more and more researches on the trajectory data mining and analysis of ship Automatic Identification System(AIS).However,the existing methods do not conduct in-depth mining and Research on the internal relationship between AIS trajectory data and port characteristics.Most of them only put forward the analysis methods for a certain feature of the port,and some of them only has a good effect on specific ports,and its applicability is low.Aiming at this problem,this thesis studies and implements the port feature analysis system based on AIS trajectory data mining.Firstly,in order to improve the efficiency of mining analysis and reduce the impact of redundant data on the analysis results,this thesis proposes a trajectory compression algorithm.Secondly,in order to provide method support for the analysis of related port characteristics,this thesis puts forward the prediction model of port monthly cargo throughput.Finally,on the basis of the above research and through the design of targeted data mining and analysis methods,this thesis puts forward the port feature analysis model,and realizes the corresponding prototype system and data visualization function.This thesis presents a trajectory compression model based on convolutional neural network.Based on the classification results,the trajectories are divided by the sub trajectories compression algorithm,and then the sub trajectories are compressed respectively.Finally,the compressed sub trajectories are optimized by the local trajectories replacement algorithm,The compression error of the trajectory is significantly reduced.In this paper,a large number of experimental data are used to test the compression effect,and the effectiveness and feasibility are verified.In this thesis,a prediction model of port monthly cargo throughput based on Deep Sets neural network is proposed.The model constructs a Deep Sets neural network model and forecasts the port monthly cargo throughput based on the characteristic sequence set composed of the deadweight tons of ships arriving at the port in the current month,the retention time at the port and the difference of draught depth between the in and out of the port.In addition,in order to improve the prediction accuracy of the model,this thesis also carries out the targeted data preprocessing operation and Feature Engineering,and optimizes the super parameters of the model based on the improved particle swarm optimization algorithm,and verifies the accuracy of the prediction by comparing with different models.This thesis further proposes a port characteristic analysis model based on AIS trajectory data mining,and designs and realizes the analysis and calculation methods of the traffic density of the road network around the port,the port connectivity index,the average retention time and average ship hour of each thousand tons of cargo loaded and unloaded in the port,the port carbon emissions,and the port monthly cargo throughput.The model realizes the analysis of the above characteristics by mining the ship AIS trajectory and other data.And combined with the relevant authoritative data results to evaluate the accuracy of the analysis results.Based on the above work,this thesis also implements the port feature analysis system based on AIS trajectory data mining,and realizes the data visualization function.The system test shows that the system can normally respond to the user’s request for port characteristics analysis,and visually display the returned analysis results including port connectivity index,monthly cargo throughput and port carbon emissions. |
