Abstract
This paper discusses the distribution regularity of ship arrival and departure and the method of prediction of ship traffic flow. Depict the frequency histograms of ships arriving to port every day and fit the curve of the frequency histograms with a variety of distribution density function by using the mathematical statistic methods based on the samples of ship-to-port statistics of Fangcheng port nearly a year. By the chi-square testing: the fitting with Negative Binomial distribution and t-Location Scale distribution are superior to normal distribution and Logistic distribution in the branch channel; the fitting with Logistic distribution is superior to normal distribution, Negative Binomial distribution and t-Location Scale distribution in main channel. Build the BP neural network and Markov model based on BP neural network model to forecast ship traffic flow of Fangcheng port. The new prediction model is superior to BP neural network model by comparing the relative residuals of predictive value, which means the new model can improve the prediction accuracy.
Highlights
Ship traffic flow is composed of ships and other water transports
In order to establish the actual vessel traffic flow characteristics model to reveal the effects of changes of regularity and other factors, it’s important to research the basic characteristics with time and space
With a given knowledge or information, the past is irrelevant to predict the future.Scholars established gray Markov chain model to predict traffic volume, which could get a better prediction accuracy to meet the requirement of short-term forecast [10,11]
Summary
Ship traffic flow is composed of ships and other water transports. In order to establish the actual vessel traffic flow characteristics model to reveal the effects of changes of regularity and other factors, it’s important to research the basic characteristics with time and space. BP artificial neural network is a network in a classical algorithm and it itself has a strong nonlinear mapping and self-learning ability to improve prediction accuracy Scholar modelled models to improve Prediction accuracy including the methods of BP (Back Propagation) neural networkoptimized BP neural network based on modified particle swarm optimization algorithm and nonlinear time series based on BP neural network [7,8,9]. With a given knowledge or information, the past (previous status) is irrelevant to predict the future (future status).Scholars established gray Markov chain model to predict traffic volume, which could get a better prediction accuracy to meet the requirement of short-term forecast [10,11]. BP neural network model and Markov chain model will be combined to improve the short-term ship traffic flow. 5) Repeatedly adjust the weighting value of each neuron, making E to achieve the error range requirements by repeat steps 3 to 5
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
