A Forecast Of Xiamen Port Throughput Based On QPSO-RBF

The prediction of port throughput is one of the six main contents of the port management planning.Optimizing the port dispatch management according to the throughput forecasting can shorten the loading and unloading waiting time,improve the efficiency of ship entry and exit,reduce the economic waste,and determine the future throughput of planning and scheduling is the key factor.This article takes Xiamen Port as an example.Firstly,it introduces the relevant theory of port decision-making,which leads to the necessity of port throughput forecast in theory and practice.Then the problem of forecasting port throughput was explored in practice and theory.Based on the traditional prediction methods,a radial basis neural network optimized by a quantum behavioral particle swarm optimization algorithm as the search algorithm is used to predict the flow of Xiamen port flow.In the process of forecasting,a method for determining the model parameters of the neural network used in combination with two different unsupervised training algorithms is proposed.That is,the combination of Density Peaks and k-means solves the implied radial basis network.The problem of confirming the number of layers of neurons,thereby improving the efficiency of neural network model construction,and avoiding the errors caused by the traditional subjective selection of neural network structure.The experimental results show that using the neural network model constructed in this paper to fit the time series data,the prediction efficiency and precision have reached a higher level.In order to cross-compare the improved neural network model used in this paper,three commonly used time series prediction models,ie,moving average method,cubic exponential smoothing method and ARIMA model are compared.The experimental results prove that the model used in this paper is compared with the above three models.The model has higher accuracy.Based on the future development of the port and production scheduling decision-making problems,this paper carries out research improvement and forecast application of the Xiamen port throughput prediction model,which has certain reference significance for real-time production port scheduling and other management decisions.