Ports operating in the same geographical range face significant competition among them. In such setting, less competitive ports may continually lose patronage of shippers （indigenous to them） to adjacent ones with b...Ports operating in the same geographical range face significant competition among them. In such setting, less competitive ports may continually lose patronage of shippers （indigenous to them） to adjacent ones with better attributes. The extent of and determinants of inter-port competition in the West Africa＇s coast are of interest to port administrators/operators who risk losing significant portion of their domestic generated cargo traffic to competing neighbouring ports. In this paper, we explore the question of what port specific attributes serve as competitive basis for West Africa＇s coastal ports operating in proximity to the other. Through a survey, users of these ports were asked to identify port specific attributes which they consider when deciding which port to use for shipments making. To enrich our empirical model, data collected from the survey were augmented with secondary data （on the identified attributes） obtained from the respective ports. Statistical evidence from data analysis suggests that ports operating in proximity in the West Africa＇s coast compete on the basis of attributes that minimise costs for port users, viz： ships＇ pre-berthiig time, ship turnround time, crane efficiency and availability of cargo spaces （proxied by frequency of ship calls）. Policy implications of the findings were discussed.展开更多
港口吞吐量时序变化数据量较小且变化快,传统长短时记忆(long short term memory,LSTM)神经网络在此类数据上易出现过拟合,导致模型预测性能不佳。针对此问题,提出融合预训练与LSTM时序模型,通过预训练捕获任务领域的全局信息,再用LSTM...港口吞吐量时序变化数据量较小且变化快,传统长短时记忆(long short term memory,LSTM)神经网络在此类数据上易出现过拟合,导致模型预测性能不佳。针对此问题,提出融合预训练与LSTM时序模型,通过预训练捕获任务领域的全局信息,再用LSTM模型精确描述各个港口的吞吐量变化规律,以提升模型对全部港口吞吐量预测的准确性。以天津港等15个中大型港口过去21年的月吞吐量为实验数据,以BP(back propagation)、自回归积分滑动平均模型(autoregressive integrated moving average model,ARIMA)、传统LSTM等预测模型和目前流行的图神经网络(graph nerual network,GNN)-LSTM模型为比较基准进行仿真实验。结果表明:所提出的融合预训练的LSTM模型能有效解决LSTM神经网络的过拟合问题,整体预测准确率高于所有基准模型。与传统LSTM模型相比,基于预训练的LSTM的MAE指标平均降低45.2%,最多降低80.0%。展开更多
文摘Ports operating in the same geographical range face significant competition among them. In such setting, less competitive ports may continually lose patronage of shippers （indigenous to them） to adjacent ones with better attributes. The extent of and determinants of inter-port competition in the West Africa＇s coast are of interest to port administrators/operators who risk losing significant portion of their domestic generated cargo traffic to competing neighbouring ports. In this paper, we explore the question of what port specific attributes serve as competitive basis for West Africa＇s coastal ports operating in proximity to the other. Through a survey, users of these ports were asked to identify port specific attributes which they consider when deciding which port to use for shipments making. To enrich our empirical model, data collected from the survey were augmented with secondary data （on the identified attributes） obtained from the respective ports. Statistical evidence from data analysis suggests that ports operating in proximity in the West Africa＇s coast compete on the basis of attributes that minimise costs for port users, viz： ships＇ pre-berthiig time, ship turnround time, crane efficiency and availability of cargo spaces （proxied by frequency of ship calls）. Policy implications of the findings were discussed.
文摘港口吞吐量时序变化数据量较小且变化快,传统长短时记忆(long short term memory,LSTM)神经网络在此类数据上易出现过拟合,导致模型预测性能不佳。针对此问题,提出融合预训练与LSTM时序模型,通过预训练捕获任务领域的全局信息,再用LSTM模型精确描述各个港口的吞吐量变化规律,以提升模型对全部港口吞吐量预测的准确性。以天津港等15个中大型港口过去21年的月吞吐量为实验数据,以BP(back propagation)、自回归积分滑动平均模型(autoregressive integrated moving average model,ARIMA)、传统LSTM等预测模型和目前流行的图神经网络(graph nerual network,GNN)-LSTM模型为比较基准进行仿真实验。结果表明:所提出的融合预训练的LSTM模型能有效解决LSTM神经网络的过拟合问题,整体预测准确率高于所有基准模型。与传统LSTM模型相比,基于预训练的LSTM的MAE指标平均降低45.2%,最多降低80.0%。
