Port Throughput Forecasting Based On Broad Learning System With Considering Multiple Influence Factors

In today's world,the economics of countries is getting more closed.The activities of commodity transportation are increasing.Because of its large volume and low freight,sea transport is the preferred mode of transport for all countries.Under such circumstances,ports play an increasingly important role.But from a sustainable perspective,the scale of port development should not exceed the market demand,or it will lead to idle equipment,resulting in the waste of resources.The construction scale must not be less than the port development needs,or the port's operation capacity will be limited.Therefore,port throughput is an important part of port planning,which can only be determined after scientific investigation and evaluation.More and more research focuses on the prediction of port throughput.The throughput of port fluctuates from time to time.It is also influenced by factors like the economic quantity of the hinterland where the port is located and the government's policy.Some forecasts suggest that all factors are time-dependent,while others suggest that factors need to be specifically quantified.The selection of throughput prediction methods will also directly affect the prediction results.In this paper,the factors affecting the port throughput are summarized.From the practical point of view,the research on port throughput by applying broad learning system with considering multiple influencing factors is proposed,which improves the prediction accuracy of throughput and fits the actual situation of port throughput prediction,making it more feasible.First of all,this paper introduces the research background and research status of port throughput at home and abroad.Then,this paper introduces the common forecasting method of port throughput in detail,and introduces its forecasting principle.Next,considering the influence factors of port throughput,this paper introduces broad learning system and other models to predict the throughput under multiple influencing factors.Then,taking Lianyun port in China as an example,based on its port throughput data from the first quarter of 2005 to the fourth quarter of 2016,this paper uses the time series method and the broad learning system to make forecasting without considering the influence factors.Furthermore,considering the impact of the second industry's investment and the third industry's investment,the broad learning system was used to predict the throughput of the port in each quarter of 2017 and 2018.Prediction experiments are also carried out by the unitary linear regression method,multiple linear regression method and BP neural network method.Those predicted results were compared with the actual throughput.In BLS,The concept of activation function enables the linear problem to be better solved in nonlinear space,and makes the training network accurate when considering the influence factors for prediction.The experimental results show that,among the above prediction methods,broad learning system has obtained the most accurate results with considering the two influencing factors at the same time,which is more suitable for the prediction of port throughput.