Research On Ship Traffic Flow Prediction Based On Multi-factor And BP Neural Network

In recent years,with the rapid development of China’s shipping industry and the improvement of the level of marine development and utilization,more and more larger ships and personnel work on the water,resulting in the number of ships and personnel in distress in the waters is also increasing year by year,coupled with the increasingly complex navigation conditions,the maritime security situation is becoming more and more serious.In this context,this thesis verifies the reliability and accuracy of the prediction model by studying the prediction method of ship traffic flow,so as to ensure the timely supervision of water safety,and achieve effective supervision of water safety and ensure the safety of life and property.At the same time,it can also provide more accurate and reasonable support and basis for the maritime management department to formulate policies and plan the port operation system.In this thesis,a new combined prediction model is established by optimizing the existing prediction model to predict the ship traffic flow.Firstly,Pearson correlation coefficient is used to analyze the correlation between ship traffic flow and various influencing factors,and then the grey correlation analysis method is used to screen out the factors with higher correlation degree as the input variables of the prediction model.Then,GA genetic algorithm and PSO standard particle swarm optimization algorithm are used as optimization algorithms to optimize the existing traditional BP neural network,and three ship traffic flow prediction models with stronger global search ability and higher prediction accuracy are constructed respectively.Taking the ship AIS data of Zhoushan Xiazhimen Channel for three consecutive months as an example,the prediction results of ship traffic flow models based on time series and correlation factors and the prediction effects of four prediction models including traditional BPNN model are analyzed respectively.The main conclusions are as follows :(1)The types of ships that mainly affect the traffic flow of the Xiazhimen Channel are working ships,container ships,bulk carriers and tugs,the length of the ship is 215 m ～ 246 m,135 m ～ 154 m,30 m ～ 49 m and 170 m ～ 194 m.(2)The accuracy of BP neural network optimized by swarm intelligence algorithm is higher than that of traditional BP neural network.The average prediction accuracy of GA-BPNN model,PSO-BPNN model and GA-PSO-BPNN model is 1.670%,2.533% and4.254% higher than that of traditional BPNN prediction model.(3)The prediction results of ship traffic flow with influencing factors is better than that of ship traffic flow only for time series.The ship traffic flow prediction based on ship type and ship length is 1.450% and 0.609% higher than the average prediction accuracy of ship traffic flow based on time series.The results show that the main influencing factors for screening and BP neural network prediction model optimized by GA-PSO hybrid algorithm can provide data foundation and technical support for ship traffic flow prediction in waterway management system.