【作者】 葛振忠；

【导师】 王洪礼；

【作者基本信息】 天津大学 ， 管理科学与工程， 2010， 博士

【摘要】 随着全球经济的快速发展,各国之间的贸易往来日益频繁,这就为物流产业的快速发展创造了更多的机遇。目前,国际贸易中90%的物资都是通过水运运输的,港口吞吐量是港口物流能力的重要表征量,也是进行各项重要决策的基础,因此对港口吞吐量进行准确的预测具有十分重要的意义。本文通过三次指数平滑法、随机梯度Boosting法、神经网络法及回归法等不同预测方法的对比,再根据误差最小的原则,最后得出“基于熵权的新陈代谢灰色马尔可夫组合预测法”对天津港吞吐量的预测具有更好的适应性结论。港口客户的忠诚度是影响港口吞吐量的最直接因素,但客户的忠诚度并不能直接观察测量到。因此,直接定量研究很少。由于客户的购买行为和忠诚度之间存在着一定的概率关系,符合隐马尔可夫过程的特征,所以,本文首先应用隐马尔可夫法仿真计算出客户忠诚度的转移规律,进而提出相应的战略措施建议确保客户的高忠诚度。影响港口吞吐量因素众多,而且这些影响因素对各港口具有不同的影响力和重要性。本文通过采用随机森林法,将腹地经济、社会物资运输量、港口集疏运能力和人口等影响因素的重要性进行排序,并从中找到最重要的影响因素,然后分析该港口影响因素的优势或制约瓶颈,并提出具体解决办法,从而实现港口快速发展的目标。研究港口物流,还必须注意到各类物资的吞吐量之间是存在着一定的相关性的。而且物资的种类不同,其装卸形式以及所需要的装卸设施设备也不同,因此,本文用研究多变量相关性的工具Copula法,研究了各类物资吞吐量的相关结构,并重点研究了天津港,最后得到了最优的GumbelCopula函数,这样既可清楚了解各类物资吞吐量的变化趋势,又可以指导对各类物资的装卸设施设备的投资强度,保证港口均衡协调的发展。更多还原

【Abstract】 With the rapid development of the global economy, increasing trade between countries create more opportunities for the rapid development of logistics industry. Currently, 90% of goods of international trade is transported through waterway. Port throughput capacity is an important symbol of the port logistics capacity, and also the basis for important decisions. So it’s of great significant of accurate forecasting of the port throughput. In this paper, according to the principle of minimum error, through comparison of various forecasting methods such as three exponential smoothing, stochastic gradient Boosting method, neural network and multivariate regression, we finally choose the "entropy-based combination of metabolic Grey Markov forecasting method" to forecast Tianjin Port throughput because of it’s effectiveness.Port customer loyalty is the most direct impact factor on port capacity, but customer loyalty can not be observed directly. Therefore, little quantitative study can be found. As the customer’s purchase behavior which can be observed directly and loyalty has certain relations of probability. It’s consistent with the characteristics of hidden Markov process, so we first use this method to calculate the transfer rules of customer loyalty, then make appropriate strategy to ensure high customer loyalty.There are many factors affecting the port throughput which have different influence. In this paper, we use RandomForest method to study the importance of factors as hinterland economy, social material traffic volume, collecting and distributing system and population. The aim is to find the most important factor, and then analyze the advantages and disadvantages in order to propose concrete solutions to achieve the goal of rapid development of the port.In studying port logistics, the correlation among different materials throughput must be attentioned. And different types of materials, require different handling facilities. This paper uses Copula as the tool of studying correlation of multivariables to study the structure of various materials throughput. As a case about Tianjin Port, the optimal Gumbel Copula function is found. This can help us understand the trend of the various materials throughput, and also guide the handling facilities investment to ensure the port keep in balance and coordinated.更多还原