基于粗糙集与遗传规划的港口吞吐量预测研究

发布时间：2018-10-31 07:07

【摘要】：中国港口吞吐量的快速增长需要港口做出适时的调整,港口吞吐量的精准预测在港口发展、改造、资源配置等方面就显得尤为重要。2008年金融危机爆发以来,我国港口吞吐量增长趋势进入了一个新的阶段,港口吞吐量增速明显变缓,多种因素都影响着吞吐量的增长,规律与以往相比发生了重大变化,传统的预测方法和模型已经难以保证当前预测工作的需要。本文研究对象为国内沿海港口的吞吐量,以广州港为代表分析港口吞吐量的影响因素,采用粗糙集与遗传规划结合的方法构建吞吐量预测模型,达到对吞吐量的精准预测,对于确定港口的发展方向、基本设施投资规模、港口的经营策略、深水泊位的选址和港口布局等方面,都有着重要的、不可缺少的作用。本文首先对以广州港为代表的中国沿海港口吞吐量影响因素进行分析,建立吞吐量指标体系；然后,采用改进的基于邻域粗糙集的数值属性约简算法获取关键指标；最后,利用遗传规划方法对关键指标样本进行训练,构建吞吐量预测模型。本文最后采用广州港2000-2012年的历史数据对模型进行实证,评价该吞吐量模型的可行性和有效性。并对广州港2013-2015年吞吐量做出预测,预测结果与广州港发展意见中2015年广州港吞吐量力争达到5亿万吨的数值基本符合,并且结合预测结果为广州港港口发展提出建议。本文的主要成果包括：1、建立了以广州港为代表的国内沿海港口的吞吐量指标体系；2、采用改进的基于邻域模型的前向贪心数值属性约简算法对港口吞吐量指标体系进行约简；3、首次将粗糙集属性约简方法与遗传规划结合用于港口吞吐量预测。
[Abstract]:The rapid growth of China's port throughput requires timely port adjustment, and accurate prediction of port throughput is particularly important in port development, transformation, and resource allocation. Since the financial crisis broke out in 2008, The growth trend of port throughput in China has entered a new stage. The growth rate of port throughput has slowed obviously, and a variety of factors have affected the growth of throughput, and the law has changed significantly compared with the past. Traditional prediction methods and models have been difficult to meet the needs of current forecasting work. The research object of this paper is the throughput of domestic coastal ports. Taking Guangzhou Port as a representative to analyze the influence factors of port throughput, a throughput prediction model is constructed by combining rough set and genetic programming to achieve accurate throughput prediction. It plays an important and indispensable role in determining the direction of port development, the scale of infrastructure investment, the management strategy of port, the location of deep water berth and the layout of port. In this paper, the factors affecting the throughput of China's coastal ports, represented by Guangzhou Port, are analyzed, and the throughput index system is established, and then the improved numerical attribute reduction algorithm based on neighborhood rough set is used to obtain the key indicators. Finally, the key index samples are trained by genetic programming method, and the throughput prediction model is constructed. In the end, we use the historical data of Guangzhou Port from 2000 to 2012 to demonstrate the feasibility and effectiveness of the model. The throughput of Guangzhou Port in 2013-2015 is forecasted, and the result is in agreement with that of Guangzhou Port in 2015, and some suggestions are put forward for the port development of Guangzhou Port. The main achievements of this paper are as follows: 1. The throughput index system of domestic coastal ports represented by Guangzhou Port is established; (2) the improved forward greedy numerical attribute reduction algorithm based on neighborhood model is used to reduce the port throughput index system; 3, the rough set attribute reduction method is combined with genetic programming for port throughput prediction for the first time.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U691

【参考文献】