| According to the United Nations Conference on Trade and Development,approximately 80% of global trade is conducted using maritime transport,so there is a clear relationship between maritime transport and the economic growth of countries.As an important part of the maritime transport system,ports are vital and play a crucial role in establishing and maintaining efficient trade routes.Good port construction can improve the competitiveness of cities and regions,attract more investment and trade,and bring more economic benefits to the country.The Yangtze River Economic Belt is an important part of China’s inland open economy system,and the ports in the Yangtze River Basin are one of the most important port clusters in China.Since its introduction in 2014,the Yangtze River Economic Belt has received great attention and support from the country.In order to achieve a high level of opening up of the Yangtze River Economic Belt to the outside world,it is necessary to give full play to the role of the Yangtze River Economic Belt as a hub for ports to link the inside and outside.Therefore,predicting specific traffic volumes becomes one of the key tasks for better planning and operation of seaports.So far,most studies have only focused on the time series of ports themselves without considering the spatial information of other ports.In this thesis,a HW-SVR spatio-temporal forecasting model is constructed by mining the spatio-temporal correlations between the series,driven by the monthly port freight volume data of 11 regions in the Yangtze River Economic Belt from January 2016 to October 2022.First of all,through data visualization,this paper briefly introduces the current situation of port freight transport in various regions of the Yangtze River Economic Belt,and conducts time series decomposition and correlation analysis on each time series.It is found that the freight volume data in various regions is seasonal and cyclical,and there is time series dependency within the time series,and there is spatial correlation between the series.Based on the above data analysis results,establish and construct a spatiotemporal prediction model that integrates temporal prediction and spatial prediction.Given the complex changes in port freight volume and their linear and nonlinear characteristics,this article selects three models: the Holt Winters model,SARIMA model,and BP neural network model to predict from a linear and nonlinear perspective,and comprehensively evaluates their fitting effects.Subsequently,this study constructed a two-dimensional SVR spatial prediction model based on mutual relationship numbers and geographic location.Based on the above research foundation,this article integrates the Holt Winters time series model with the most ideal fitting effect and the SVR spatial prediction model through stacking to construct the HW-SVR spatiotemporal prediction model.This model not only considers the characteristics of the time series itself,but also takes into account the spatial correlation between ports.Whether compared with three different time series models or SVR spatial models,the prediction accuracy of the HW-SVR model has significantly improved under the comparison of three different evaluation indicators.In addition,when faced with significant irregular changes in freight volume in the short term,the model can still maintain relatively reasonable prediction results,thus proving the effectiveness of the model.This study used this model to predict the monthly port freight volume of various regions in the Yangtze River Economic Belt in the next year. |
