遗传算法优化的BP神经网络在连云港港口吞吐量预测中的应用研究

发布时间：2018-04-10 18:51

本文选题：吞吐量预测 + 影响因素　；参考：《深圳大学》2017年硕士论文

【摘要】：近年来,受益于中国经济的日益繁荣,中国港口已经成为世界港口中至关重要的组成部分,而作为反映港口发展状况的一个最基本的指标,港口吞吐量的相关研究也多种多样。其中港口吞吐量的预测对于港口发展规划有着重要的意义。然而,目前的港口吞吐量预测方法仍然存在一些不足,传统的预测方法多采用时间序列或只选取一个因素参与预测,并不能充分体现出港口腹地的经济发展、季节条件等多种因素对港口吞吐量的影响。因此,本文将探究多种因素与港口吞吐量之间的联系,并将影响因素数据用于接下来的连云港港吞吐量的预测。首先,本文将从宏观和微观两大方面入手分析这些因素对港口吞吐量的影响,并从有限的可以获得的数据中确定了季节性因素对吞吐量有相关性,因此本文将按照季度预测吞吐量并且会将去年当季和前年当季吞吐量作为BP神经网络输入数据;同时发现二季度前第二产业和第三产业固定资产投资与连云港港口吞吐量影响因素具有较大的相关系数,可以用于港口吞吐量预测。接着介绍多种传统的吞吐量预测方法,并从中选取了三次指数平滑法和一元回归预测法两种方法,以连云港港作为实证研究对象,对连云港港口吞吐量进行了预测,并分析了预测方法存在的缺点。接着,为了克服传统预测方法存在的这些问题,本文在MATLAB中构建BP神经网络,选取前面选取的连云港港口吞吐量影响因素数据作为实证研究对象进行预测,发现预测结果并不理想,通过对预测结果进行分析,发现是BP神经网络中的初始权值和阈值的随意性影响了连云港港口吞吐量预测结果。所以本文选择了遗传算法对BP神经网络进行优化,以求得到更加理想的预测结果。通过遗传算法处理得到的最优初始权值和阈值,并将初始权值和阈值赋值给BP神经网络,并实现网络的训练以及预测输出,得到相关的预测结果,把新得出的评价结果和之前的评价结果进行对比分析可以发现,经过遗传算法优化权值和阈值后的BP神经网络进行评价的效果更佳。证明了遗传算法优化的BP神经网络在连云港港口吞吐量预测过程中的可行性和有效应。
[Abstract]:In recent years, thanks to the increasing prosperity of China's economy, Chinese ports have become a vital part of the world's ports, and as a most basic indicator of port development, port throughput research is also diverse.The prediction of port throughput is of great significance to port development planning.However, there are still some shortcomings in the current forecasting methods of port throughput. The traditional forecasting methods usually use time series or select only one factor to participate in the prediction, which can not fully reflect the economic development of the port hinterland.The influence of seasonal conditions and other factors on port throughput.Therefore, this paper will explore the relationship between various factors and port throughput, and use the data of influencing factors to predict the throughput of Lianyungang port.First of all, this paper will analyze the impact of these factors on port throughput from macro and micro aspects, and determine the correlation between seasonal factors and throughput from the limited data available.Therefore, this paper will forecast the throughput according to the quarter and will take the throughput of last year and the year before as the input data of BP neural network.At the same time, it is found that the fixed assets investment of the secondary industry and the tertiary industry have a large correlation coefficient with the port throughput of Lianyungang before the second quarter, which can be used to forecast the port throughput.Then several traditional throughput forecasting methods are introduced and two methods of cubic exponential smoothing and unitary regression forecasting are selected to predict the throughput of Lianyungang port.The shortcomings of the prediction method are also analyzed.Then, in order to overcome these problems of traditional forecasting methods, this paper constructs BP neural network in MATLAB, and selects the data of the influence factors of port throughput of Lianyungang port as the empirical research object to forecast.It is found that the prediction results are not ideal. By analyzing the prediction results, it is found that the randomness of the initial weights and thresholds in the BP neural network affects the forecasting results of the port throughput of Lianyungang port.Therefore, genetic algorithm is selected to optimize BP neural network to obtain more ideal prediction results.The optimal initial weights and thresholds are processed by genetic algorithm, and the initial weights and thresholds are assigned to BP neural networks.By comparing the new evaluation results with the previous ones, it can be found that the BP neural network after optimizing the weights and thresholds of genetic algorithm is more effective.It is proved that the BP neural network optimized by genetic algorithm is feasible and effective in forecasting the throughput of Lianyungang port.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U652.14;TP183

【相似文献】